Calmodulin (CaM) regulation of the CaV1–2 family of Ca2+ channels ranks among the most consequential of biological Ca2+ decoding systems1,2. In this regulation, the Ca2+-free form of CaM (apoCaM) already pre-associates with channels3,4,5, ready for ensuing Ca2+-driven modulation of channel opening. Upon elevation, intracellular Ca2+ binds to this indwelling CaM, driving conformational changes that enhance opening in some channels6,7,8 (positive-feedback ‘facilitation’) and inhibit opening in others9,10 (negative-feedback ‘inactivation’). Intriguingly, Ca2+ binding to the individual carboxy- and amino-terminal lobes of CaM can semiautonomously induce distinct components of channel regulation7,9,11, where the C-lobe responds well to Ca2+ entering through the channel on which the corresponding CaM resides (‘local Ca2+ selectivity’), and the N-lobe may, in some channels, require the far weaker Ca2+ signal from distant Ca2+ sources6,7,12,13,14 (‘global Ca2+ selectivity’). Such Ca2+-feedback regulation influences many biological functions1,15,16,17 and furnishes mechanistic lessons for Ca2+ decoding14. Indeed, CaM regulation of L-type (CaV1.2) channels strongly influences cardiac electrical stability15,18, and pharmacological manipulation of such regulation looms as a future antiarrhythmic strategy18,19.

Crucial for understanding and manipulating this CaM regulatory system is the identification of the conformations that underlie such Ca2+ modulation. Figure 1a summarizes the currently accepted conceptual framework, with specific reference to L-type CaV1.3 channels for concreteness. Configuration E (‘empty’ of CaM) represents channels lacking preassociated apoCaM. Such channels can open normally, but do not exhibit Ca2+/CaM-dependent inactivation (CDI) over the typical ~300 ms duration of channel-activity measurements20. Over this period, Ca2+/CaM from the bulk solution cannot appreciably access a channel in configuration E to produce CDI20,21,22,23. ApoCaM preassociation with configuration E yields channels in configuration A, where opening can also proceed normally, but subsequent CDI can now ensue. A thereby denotes channels that are ‘active’ and capable of CDI. Switching between configurations E and A occurs slowly (>10 s of seconds24), and almost exclusively involves apoCaM, because typical experiments only briefly activate Ca2+ channels every 20–30 s. Thus, there is negligible exchange with configuration E during typical measurements of current. Regarding CDI, Ca2+ binding to both lobes of CaM yields configuration ICN (both C- and N-lobes of CaM engaged towards CDI), corresponding to a fully inactivated channel with strongly reduced opening2,25. As for intermediate configurations7,9,11,14,26, Ca2+ binding only to the C-lobe induces configuration IC, representing a C-lobe inactivated channel with reduced opening; Ca2+ binding only to the N-lobe yields an analogous N-lobe-inactivated configuration (IN), also with reduced opening. Subsequent entry into configuration ICN likely involves cooperative interactions denoted by a λ symbol. Overall, CDI reflects redistribution from configuration A into IC, IN and ICN. Of note, we exclude cases where one Ca2+ binds a lobe of CaM, because binding within lobes is highly cooperative27. Moreover, only one CaM is included, based on multiple lines of evidence22,23.

Figure 1: General schema for CaM regulation of representative L-type CaV1.3 channels.
figure 1

(a) Primary configurations of CaM/channel complex with respect to CaM-regulatory phenomena (E, A, IC, IN and ICN). Inset at far right, cartoon of main channel landmarks involved in CaM regulation, with only the pore-forming α1D subunit of CaV1.3 diagrammed. Ca2+-inactivation (CI) region, in the proximal channel C terminus (~160 amino acids (aa)), contains elements potentially involved in CaM regulation. IQ domain (IQ), comprising the C-terminal ~30 aa of the CI segment, long proposed as preeminent for CaM/channel binding. Dual vestigial EF-hand (EF) motifs span the proximal ~100 aa of the CI module; these have been proposed to play a transduction role in channel regulation. Proximal CI (PCI) region constitutes the CI element exclusive of the IQ domain. NSCaTE on channel N terminus of CaV1.2–1.3 channels may be the N-lobe Ca2+/CaM effector site. (b) Whole-cell CaV1.3 currents expressed in HEK293 cell, demonstrating CDI in the presence of endogenous CaM only. CDI observed here can reflect properties of the entire system diagrammed in a, as schematized by the stick-figure diagram at the bottom of b. Here and throughout, the vertical scale bar pertains to 0.2 nA of Ca2+ current (black); and the Ba2+ current (gray) has been scaled ~3-fold downward to aid comparison of decay kinetics, here and throughout. Horizonal scale bar, 100 ms. (c) Currents during overexpression of CaMWT, isolating the behaviour of the diamond-shaped subsystem at bottom. (d) Currents during overexpression of CaM12, isolating C-lobe form of CDI. (e) Currents during overexpression of CaM34, isolating N-lobe form of CDI. (ce) Vertical bar, 0.2 nA Ca2+ current. Timebase as in b.

The structural basis of this conceptual foundation is less certain, but has been dominated by an IQ-centric hypothesis, where an IQ domain, present on the C termini of all CaV1–2 channels2 (Fig. 1a, far right, blue circle), serves as the dominant CaM-binding locus on the channel. By this hypothesis, not only does this element comprise much of the preassociation surface for apoCaM4,5,20 (Fig. 1a, configuration A), it also constitutes the primary effector site2,5,7,9,10,25,28 for Ca2+/CaM rebinding to induce Ca2+ regulation (for example, Fig. 1a, ICN). The predominance of the IQ-centric paradigm2 has prompted resolution of several crystal structures of Ca2+/CaM complexed with IQ-domain peptides of various CaV1–2 channels29,30,31,32.

Nonetheless, certain findings fit poorly with this viewpoint. First, crystal structures of Ca2+/CaM complexed with wild-type (WT) and mutant IQ peptides of CaV1.2 indicate that a signature isoleucine in the IQ element is deeply buried within the C-lobe of Ca2+/CaM, and that alanine substitution at this isoleucine negligibly perturbs structure30. Moreover, Ca2+/CaM affinities for analogous WT and mutant IQ peptides are nearly identical28. How then does alanine substitution at this well-encapsulated locus influence the rest of the channel to strongly disrupt functional regulation30? Second, in CaV1.2/1.3 channels we have demonstrated that the effector interface for the N-lobe of Ca2+/CaM resides within an N-terminal spatial Ca2+ transforming element (NSCaTE) element of the channel N terminus13,14,33 (Fig. 1a, far right), separate from the IQ element. Third, analysis of the atomic structure of Ca2+/CaM bound to an IQ peptide of CaV2.1 channels hints that the C-lobe effector site also resides somewhere outside the IQ module31. In all, the long disconnect between challenges like these and IQ-centric theory represents a critical impasse in the field.

A major concern with prior IQ-domain analyses is that function was mostly characterized with only endogenous CaM present5,10,25,28,31. This regime is problematic, as IQ-domain mutations could alter CaM regulation via perturbations at multiple steps within Fig. 1a, whereas interpretations mainly ascribe effects to altered Ca2+/CaM binding with an IQ effector site. Serious interpretive challenges thus include the following: (1) although the high apoCaM affinity of most WT channels4,20 renders configuration E unlikely (Fig. 1a), this may not hold true for mutant channels, just as observed for certain CaV1.3 splice variants20. Mutations weakening apoCaM preassociation could thereby reduce CDI by favouring configuration E (Fig. 1a, incapable of CDI), without affecting Ca2+/CaM binding. (2) Mutations that do weaken interaction with one lobe of Ca2+/CaM may have their functional effects masked by cooperative steps (λ in Fig. 1a).

This study systematically investigates the IQ-centric hypothesis, minimizing the above challenges by focusing on CaV1.3 channels, a representative L-type channel whose CDI is particularly robust and separable into distinct C- and N-lobe components11,13,14. These attributes simplify analysis as follows. For orientation, Fig. 1b illustrates the CDI of CaV1.3 channels expressed in HEK293 cells, with only endogenous CaM present. Strong CDI is evident from the rapid decay of whole-cell Ca2+ current (black trace) compared with the nearly absent decline of Ba2+ current (gray trace). As Ba2+ binds negligibly to CaM34, the fractional decline of Ca2+ versus Ba2+ current after 300 ms depolarization quantifies the steady-state extent of CDI (Fig. 1b, right, CDI parameter). The CDI here reflects the operation of the entire Fig. 1a system, as schematized at the bottom of Fig. 1b. We can formally isolate the diamond-shaped subsystem lacking configuration E (Fig. 1c, bottom), by using mass action and strong overexpression of WT CaM (CaMWT). The resulting CDI (Fig. 1c) is indistinguishable from that with only endogenous CaM present (Fig. 1b), owing to the high apoCaM affinity of WT CaV1.3 channels. Full deconstruction of CDI arises upon strong coexpression of channels with a mutant CaM that only allows Ca2+ binding to its C-terminal lobe9 (Fig. 1d, CaM12). With reference to Fig. 1a, this manoeuvre depopulates configuration E by mass action, and forbids access into configurations IN and ICN. Thus, the isolated C-lobe component of CDI11,14 is resolved (Fig. 1d), with its signature rapid timecourse of current decay. Importantly, this regime avoids interplay with cooperative λ steps in Fig. 1a. Likewise, strongly coexpressing mutant CaM exhibiting Ca2+ binding to its N-lobe alone9 (CaM34) isolates the slower N-lobe form of CDI11,14 (Fig. 1e), with attendant simplifications. Thus armed, we here exploit selective monitoring of CaV1.3 subsystems (Fig. 1b–e), combined with alanine, scanning mutagenesis of the entire carboxyl tail of CaV1.3 channels. In doing so, we argue against the IQ-centric paradigm and propose a new framework for the CaM regulation of Ca2+ channels.


Individually transformed Langmuir analysis of CaM/channel regulation

Identifying channel effector interfaces for Ca2+/CaM is challenging. The main subunit of CaV channels alone spans about 2,000 amino acids or more; and peptide assays indicate that Ca2+/CaM can bind to multiple segments of uncertain function25,35,36,37,38,39. Even if mutating these segments alters CaM regulation, the observed functional effects could reflect perturbations of apoCaM preassociation, Ca2+/CaM binding or transduction. To address these challenges, we initially consider an expanded conceptual layout believed valid for either isolated N- or C-lobe CDI14 (Fig. 2a), then deduce from this arrangement a simple quantitative analysis to identify bona fide effector interfaces. An apoCaM lobe begins prebound to a channel preassociation surface (state 1). Ca2+ binding to CaM in this prebound state is considered rare14,40. However, after apoCaM releases (state 2), it may bind Ca2+ to produce Ca2+/CaM (state 3), or return to state 1. The transiently dissociated lobe of CaM (state 2 or 3) remains within a channel alcove over the usual timescale of CaM regulation (≤seconds). Finally, Ca2+/CaM binds a channel effector site (state 4, square pocket), ultimately inducing regulation via transduction to state 5. Emergent behaviours of this scheme rationalize local and global Ca2+ selectivities, as argued previously14.

Figure 2: Probing functionally relevant CaM regulatory interactions via iTL analysis.
figure 2

(a) Isolated C- or N-lobe regulatory system (denoted by stick-figure diagrams on left) can be coarsely represented by a five-state scheme on right. A single lobe of apoCaM begins preassociated to channel (state 1). Following disassociation (state 2), CaM may bind two Ca2+ ions (state 3, black dots). Ca2+/CaM may subsequently bind to channel effector site (state 4). From here, transduction step leads to state 5, equivalent to CDI. Association constant for lobe of apoCaM binding to preassociation site is ε; whereas γ1 and γ2 are association constants for respective transitions from states 3 to 4, and states 4 to 5. (b) Unique Langmuir relation (equation 1) that will emerge upon plotting channel CDI (defined Fig. 1b, right) as a function of Ka,EFF (association constant measured for isolated channel peptide), if Ka,EFF is proportional to one of the actual association constants in the scheme as in a. Black symbols, hypothetical results for various channel/peptide mutations; green symbol, hypothetical WT. (c) Predicted outcome if peptide association constant Ka,EFF has no bearing on association constants within holochannels. (d) Outcome if mutations affect holochannel association constants, but not peptide association constants. (e) Outcome if mutations affect holochannel association constant(s) and peptide association constant, but in ways that are poorly correlated.

Despite the multiple transitions present even for this reduced CDI subsystem (Fig. 2a, left schematics), a straightforward relationship emerges that will aid detection of Ca2+/CaM interfaces on the channel, as follows. Suppose we can introduce point alanine mutations into the channel that selectively perturb the Ca2+/CaM-binding equilibrium association constant γ1 (Fig. 2a). Moreover, suppose we can measure Ca2+/CaM binding to a corresponding channel peptide, and the affiliated association constant ϰa,EFF is proportional to γ1 in the channel. It then turns out that our metric of inactivation (CDI in Fig. 1b) will always be given by the Langmuir function

where CDImax is the value of CDI if ϰa,EFF becomes exceedingly large, and Λ is a constant comprised of other association constants in the layout (Supplementary Note S1). Figure 2b plots this function, where the green symbol marks a hypothetical WT channel position, and mutations should create data symbols that decorate the remainder of the curve. Importantly, the requirement that peptide Ka,EFF need only be proportional to (not equal to) holochannel γ1 increases the chances that tagged peptides may suffice to correlate with holochannel function. In addition, equation 1 will hold true only if these two suppositions are satisfied (Supplementary Note S2). For example, if mutations alter two transitions within the holochannel, a function with different shape will result. Alternatively, if mutations change the peptide interaction with Ca2+/CaM (ϰa,EFF), but not any of the actual association constants within the channel, the outcome in Fig. 2c will emerge. In this case, though the channel peptide can bind Ca2+/CaM in isolation, this reaction has no bearing on transitions within the intact holochannel (Fig. 2a). By contrast, Fig. 2d diagrams a scenario where mutations actually do affect transition(s) governing CDI within the holochannel, yet altogether fail to perturb Ca2+/CaM binding to a peptide segment of the channel. It is also possible that mutations could affect transition(s) governing CDI within the holochannel, but in ways that are uncorrelated with mutational perturbations of Ca2+/CaM binding to a corresponding peptide segment (Fig. 2e). The red symbol denotes a specific subset of this scenario, where a mutation affects transition(s) within the holochannel so as to enhance CDI, whereas the same mutation produces uncorrelated diminution of Ca2+/CaM binding to a peptide segment of the channel. Yet other deviations from equation 1 are possible, including those arising from the existence of effector sites beyond our alanine scan (Supplementary Note S3). Importantly, these outcomes will pertain, regardless the size and complexity of the scheme in Fig. 2a (Supplementary Note S4). Because of this generality, we term the analysis individually transformed Langmuir (iTL) analysis.

Given this insight, we undertook alanine-scanning mutagenesis of CaV1.3 channel domains, and screened electrophysiologically for altered CaM regulatory hotspots. In parallel, we introduced hotspot mutations into peptides overlapping scanned regions, and estimated Ka,EFF of potential CaM binding. For this purpose, we utilized live-cell fluorescence resonance energy transfer (FRET) two-hybrid assays3,4,41, which have the resolution and throughput for the task. If such binding truly reflects holochannel function, then CDI should vary with Ka,EFF as a Langmuir function (equation 1, Fig. 2b). By contrast, if Ka,EFF changes in a manner unrelated to holochannel CDI, data would diverge from equation 1 (Fig. 2c–e, or otherwise). CaM effector interfaces could thus be systematically resolved.

iTL analysis of IQ domain as Ca2+/CaM effector site

We first addressed whether the CaV1.3 IQ domain serves as a Ca2+/CaM effector site for CDI, as IQ-centric theory postulates. Single alanines were substituted at each position of the entire IQ domain of CaV1.3 channels, whose sequence appears atop Fig. 3a, with the signature isoleucine bolded at position ‘0’. Naturally occurring alanines were changed to threonine. CDI of these mutants was then characterized for the isolated N- and C-lobe CDI subsystems described above (Fig. 3a, left schematics), thus minimizing potential complications from diminished preassociation with apoCaM (Fig. 1a, configuration E), or masking of CDI effects by cooperative λ steps (Fig. 1a). Whereas little deficit in N-lobe CDI was observed (Fig. 3a), the C-lobe CDI was strongly attenuated by alanine substitutions at I[0]A (Fig. 3b, red bar, exemplar traces) and nearby positions (rose). To test for correspondence between reductions in C-lobe CDI and altered Ca2+/CaM binding, we performed FRET two-hybrid assays of Ca2+/CaM binding to alanine-substituted IQ peptides, with substitutions encompassing sites associated with the strongest CDI effects (Fig. 3b, red and rose bars). Hatched bars denote additional sites chosen at random. The left aspect of Fig. 3c cartoons the FRET interaction partners, and the right portion displays the resulting binding curve for the WT IQ peptide (Fig. 3c, right, black). FR-1 is proportional to FRET efficiency, as indicated by the efficiency EA scale bar on the right. Dfree is the free concentration of donor-tagged molecules (cyan fluorescent protein (CFP)–CaM), where 200 nM is approximately 6,100 Dfree units4,42. At odds with a Ca2+/CaM effector role of the IQ domain, the binding curve for the I[0]A substitution (Fig. 3c, right, red) resembled that for the WT peptide (black), whereas C-lobe CDI was strongly decreased (Fig. 3b). Figure 3c (middle) displays a bar-graph summary of the resulting association constants (Ka,EFF); the WT value is shown as a dashed green line, and that for I[0]A as a red bar (Supplementary Note S5). If the IQ domain were the effector site for the C-lobe of Ca2+/CaM, C-lobe CDI over various substitutions should correlate with association constants according to equation 1 (Supplementary Notes S1 and S6). However, plots of our data markedly deviate from such a relation (Fig. 3e), much as in Fig. 2e. The green symbol denotes the WT IQ case. Likewise, plots of N-lobe CDI versus Ka,EFF deviated from a Langmuir (Fig. 3d), much as in Fig. 2c. These outcomes fail to support the IQ domain as an effector site for either lobe of Ca2+/CaM. The actual role of the IQ domain in CDI will be explored later in Fig. 6.

Figure 3: Inconsistencies with IQ domain role as Ca2+/CaM effector site.
figure 3

(a) No appreciable deficit in isolated N-lobe CDI upon point alanine substitutions across the IQ domain (sequence at top with bolded isoleucine at ‘0’ position). Left, corresponding subsystem schematic. Middle, bar-graph summary of CDI metric, as defined in Fig. 1b. Bars, mean±s.e.m. for ~6 cells each. Green dashed line, WT profile; red bar, I[0]A; blue symbol in all panels, Y[3]D. Right, exemplar currents, demonstrating no change in N-lobe CDI upon I[0]A substitution. Horizontal scale bar, 100 ms; vertical scale bar, 0.2 nA Ca2+ current. Red, Ca2+ current; gray, Ba2+ current. (b) Isolated C-lobe CDI (corresponding subsystem schematized on left) exhibits significant attenuation by mutations surrounding the central isoleucine (coloured bars). Format as in a. I[0]A shows the strongest attenuation (red bar and exemplar currents at right). Bars average ~5 cells±s.e.m. Dashed green line, WT profile. Timebase as in b; vertical scale bar, 0.2 nA Ca2+ current. (c) Bar-graph summary of association constants (Ka,EFF=1/Kd,EFF) for Ca2+/CaM binding to IQ, evaluated for constructs exhibiting significant effects in b (coloured bars, with I[0]A in red), or chosen at random (hashed in b). Error bars, non-linear s.d. estimates. FRET partners schematized on the left, and exemplar binding curves on the right for I[0]A (red) and WT (black). Symbols average ~7 cells. Smooth curve fits, 1:1 binding model. Calibration to efficiency EA=0.1, far right vertical scale bar. Horizontal scale bar corresponds to 100 nM. (d) Plots of N-lobe CDI versus Ka,EFF deviate from equation 1, much as in Fig. 2c. Green, WT; red, I[0]A; blue, Y[3]D. (e) Plots of C-lobe CDI versus Ka,EFF also diverge from Langmuir, as in Fig. 1e. This result further argues against the IQ per se acting as an effector site for the C-lobe of Ca2+/CaM. Symbols as in d. (d,e) Y[3]D (blue symbol, CDI mean of four cells) yields poor Ca2+/CaM binding, but unchanged CDI. Supplementary Note S7, further FRET data.

To undertake a still more stringent test, we investigated a Y[3]D construct, based on a prior analogous mutation in CaV2.1 that intensely diminished Ca2+/CaM affinity31. Indeed, the Y[3]D substitution in CaV1.3 resulted in a large 13.5-fold decrement in Ka,EFF (Fig. 3c, blue symbol). However, there was no change in C- or N-lobe CDI (Fig. 3a, blue symbols; Supplementary Note S7). These data deviated yet more strongly from a Langmuir (blue symbols, Fig. 3d), arguing further against the IQ domain as a Ca2+/CaM effector site.

NSCaTE element upheld as effector site for N-lobe of Ca2+/CaM

Given the absence of a positive outcome for iTL analysis of the IQ domain (that is, Fig. 2b), we turned to the N-terminal NSCaTE module (Fig. 4a, oval), previously proposed as an effector site for N-lobe CDI13,14. For reference, Fig. 4b displays the WT CaV1.3 profile for N-lobe CDI. Single alanines were substituted across the NSCaTE module (Fig. 4d, top), at residues that were not originally alanine. The bar-graph summary below (Fig. 4d) indicates strongly diminished N-lobe CDI upon alanine substitution at three residues, previously identified as critical13,14 (W[44]A, I[48]A and R[52]A). For comparison, the WT level of CDI is represented by the green dashed line and affiliated error bars. W[44]A featured the strongest CDI decrement, as shown by the Ca2+ current (Fig. 4b, red trace) and population data (Fig. 4d, red bar). To pursue iTL analysis, we characterized corresponding binding curves between NSCaTE and Ca2+/CaM34 FRET pairs (Fig. 4c, left; Supplementary Note S8). The WT pairing exhibited a well-resolved binding curve with Ka,EFF=4 × 10−4 Dfree−1 units (Fig. 4c, right, black), whereas the W[44]A variant yielded a far lower affinity with Ka,EFF~0 (red). A summary of binding affinities is shown for this and additional mutations within NSCaTE in Fig. 4e (Supplementary Note S9), where the dashed green line signifies the WT profile. The crucial test arises by plotting N-lobe CDI as a function of Ka,EFF, which resolves the Langmuir relation in Fig. 4f. For reference, WT is shown in green and W[44]A in red. The particular formulation of equation 1 for this arrangement is given in Supplementary Note S1. Hence, iTL analysis does uphold NSCaTE as a predominate effector site for N-lobe CDI, as argued before by other means13,14. By contrast, analysis of C-lobe CDI (Fig. 4g–k and Supplementary Note S10) reveals deviation from equation 1 (Fig. 4l), much as in Fig. 2c. Thus, NSCaTE mutations have little bearing on C-lobe CDI of the holochannel, though such mutations affect Ca2+/CaM12 binding to an isolated NSCaTE peptide.

Figure 4: iTL analysis of Ca2+/CaM effector role of NSCaTE module of CaV1.3 channels.
figure 4

(a) Cartoon depicting NSCaTE as putative effector interface for N-lobe of Ca2+/CaM. (b) Exemplar CaV1.3 whole-cell currents exhibiting robust isolated N-lobe CDI, as seen from the rapid decay of Ca2+ current (black trace). Corresponding stick-figure subsystem appears on the left. W[44]A mutation abolishes N-lobe CDI, as seen from the lack of appreciable Ca2+ current decay (red trace). Gray trace, averaged Ba2+ trace for WT and W[44]A constructs. Horizontal scale bar, 100 ms; vertical scale bar, 0.2 nA Ca2+ current (red, W44A; black, WT). (c) FRET two-hybrid binding curves for Ca2+/CaM34 and NSCaTE segment, with FRET partners schematized on the left. WT pairing in black; W[44]A mutant pairing in red. Each symbol, mean±s.e.m. of ~5 cells. (d) Bar-graph summary of N-lobe CDI for NSCaTE mutations measured after 800 ms depolarization, with NSCaTE sequence at the top, as numbered by position within CaV1.3. Data for W[44]A in red; dashed green line, WT. Bars, mean±s.e.m. of ~5 cells. (e) Association constants (Ka,EFF=1/ Kd,EFF) for Ca2+/CaM34 binding to NSCaTE module evaluated for constructs exhibiting significant effects in panel d. Error bars, non-linear s.d. estimates. Data for W[44]A in red; dashed green line, WT. (f) Plotting N-lobe CDI versus Ka,EFF uncovers a Langmuir, identifying NSCaTE as functionally relevant effector site. W[44]A in red; WT in green. (gl) iTL fails to uphold NSCaTE as effector site for C-lobe of Ca2+/CaM. Format as in af. (h,j) C-lobe CDI at 300 ms, unchanged by NSCaTE mutations. Bars in j, mean±s.e.m. of ~5 cells. (i,k) Changes in Ka,EFF of NSCaTE module for Ca2+/CaM12 via 33-FRET. Each symbol in i, mean±s.e.m. of ~5 cells. (l) C-lobe CDI versus Ka,EFF deviates from Langmuir, as in Fig. 2c.

Identification of the C-lobe Ca2+/CaM effector interface

Satisfied by proof-of-principle tests of the iTL approach, we turned to identification of the as-yet-unknown effector site for the C-lobe form of CDI. Our screen focused upon the entire carboxy tail of CaV1.3 channels upstream of the IQ domain (Fig. 5a, proximal Ca2+-inactivation (PCI) domain), because switching these C-terminal segments in chimeric channels sharply influences this type of CDI31,43. For completeness, we initially characterized isolated N-lobe CDI for mutations throughout the PCI and found no appreciable decrement from WT levels (Fig. 5b and Supplementary Note S12) as expected. Gaps indicate non-expressing configurations. By contrast, for isolated C-lobe CDI, the sharp diminution of CDI upon LGF→AAA substitution (Fig. 5c, red) exemplifies just one of many newly discovered ‘hotspot’ loci residing in the PCI midsection (Fig. 5f, rose and red; Supplementary Note S12). As a prelude to iTL analysis, we determined the binding of Ca2+/CaM to the PCI element (Fig. 5g, left cartoon), and indeed the LGF substitution weakens interaction affinity (Fig. 5d). Likewise, binding of the isolated C-lobe of Ca2+/CaM to PCI was similarly attenuated by the LGF mutation (Supplementary Note S11), arguing explicitly for disruption of a C-lobe interface. In addition, for loci demonstrating the strongest reduction in C-lobe CDI (Fig. 5f, rose and red), corresponding Ca2+/CaM affinities were determined to also attenuate Ka,EFF (Fig. 5g and Supplementary Notes S11 and S12). Importantly, graphing C-lobe CDI versus binding affinity strikingly resolves a Langmuir relation (Fig. 5i), furnishing compelling evidence that the PCI midsection comprises an effector interface for the C-lobe of Ca2+/CaM. The green symbol corresponds to WT, and the red datum to the LGF mutant. Supplementary Note S1 specifically formulates equation 1 for this case. As expected, plots of N-lobe CDI versus binding affinity deviate from a Langmuir (Fig. 5h), much as in Fig. 2c. Overall, the impressive mirror-like inversion of results for NSCaTE (Fig. 4f) and PCI (Fig. 5h) underscores the considerable ability of iTL analysis to distinguish between effector sites of respective N- and C-lobe CDI.

Figure 5: iTL analysis of PCI segment as C-lobe Ca2+/CaM effector interface.
figure 5

(a) Channel cartoon depicting PCI segment as putative effector site for C-lobe of Ca2+/CaM. (b) Isolated N-lobe CDI for WT and LGF→AAA (LGF) mutant channels. Ca2+ current for WT in red, and for LGF in red. Gray, averaged Ba2+ trace. Horizontal scale bar, 100 ms; vertical scale bar, 0.2 nA Ca2+ current (red, LGF; black, WT). (c) Isolated C-lobe CDI for WT and LGF mutant channels, indicating strong CDI attenuation by LGF mutation. Format as in b. (d) FRET two-hybrid binding curves for Ca2+/CaM pitted against PCI segments, for WT (black) and LGF (red). Each symbol, mean±s.e.m. from~9 cells. (e) Bar-graph summary confirming no appreciable reduction of isolated N-lobe CDI, over all alanine scanning mutants across the PCI region (sequence at the top). Schematic of corresponding system under investigation at the left. Green dashed line, WT; red, LGF mutant; gaps, non-expressing configurations. Bars, mean±s.e.m. of ~5 cells. (f) Bar-graph summary, C-lobe CDI for alanine scan of PCI. Red bar, LGF→AAA mutant showing strong CDI reduction. Rose bar, other loci showing substantial CDI reduction. Hashed, randomly chosen loci for subsequent FRET analysis below. Bars, mean±s.e.m. of ~5 cells. (e,f) CDI decrease for YLT cluster (Fig. 5e) reflects reduced Ca2+ entry from 30 mV depolarizing shift in activation, not CDI attenuation per se. Shifts for all other loci were at most ±10 mV (not shown). (g) Association constants for Ca2+/CaM binding to PCI region, with FRET partners as diagrammed on the left. Green dashed line, WT profile. PCI mutations yielding large C-lobe CDI deficits were chosen for FRET analysis (red and rose in f), as well as those chosen at random (hashed in f). Error bars, nonlinear estimates of standard deviation. (h) Plots of N-lobe CDI versus Ka,EFF for Ca2+/CaM binding to PCI deviated from Langmuir. Red, LGF; green, WT. (i) Alternatively, plotting C-lobe CDI revealed Langmuir relation, supporting PCI as C-lobe Ca2+/CaM effector site. Symbols as in h.

C-lobe CDI also requires IQ domain interaction with the PCI element

Though the IQ domain alone does not appear to be an effector site for Ca2+/CaM (Fig. 3), alanine substitutions in this element nonetheless attenuated the C-lobe CDI7,10,11,28,31, a result reproduced for reference in Fig. 6a. Might the departure of Ca2+/CaM to NSCaTE (Fig. 4) and PCI elements (Fig. 5) then allow the IQ domain to rebind elsewhere, in a manner also required for C-lobe CDI? Thus viewed, IQ-domain mutations could diminish C-lobe CDI by weakening this rebinding, but in a way that correlates poorly with IQ-peptide binding to Ca2+/CaM in isolation. As C-lobe CDI can be conferred to CaV2 channels by substituting PCI and IQ elements from CaV131,43, will the requisite rebinding involve association between these very elements?

Figure 6: Role of IQ domain in C-lobe CDI.
figure 6

(a) Cartoon depicting putative binding interaction between IQ domain and PCI segment, which is also required for C-lobe CDI. (b) Bar-graph summary of C-lobe CDI measured for alanine scan of IQ domain, reproduced from Fig. 3b. Strongest CDI reduction for I[0]A mutant (red), followed closely by loci affiliated with rose and blue bars underneath black dashed line. Dashed green line, WT. (c) Association constants Ka,EFF determined for 33-FRET binding between IQ domain and PCI region (partners diagrammed at left), under elevated levels of Ca2+. WT profile, green dashed line. Bars, Ka,EFF for mutants with strongest effects (colored bars in b) or chosen at random (hashed bars in b). Error bars, nonlinear s.d. estimates. (d) Exemplar 33-FRET binding curves for IQ/PCI interaction. Each symbol, mean±s.e.m. of ~8 cells. Absent Ca2+, the IQ domain associates only weakly with the PCI region (gray). However, elevated Ca2+ greatly enhances binding (black). (e) 33-FRET binding curves for I[0]A (red) and Q[1]A (blue) mutations under elevated Ca2+. Each symbol, mean±s.e.m. of ~5 cells. Fit for WT IQ/PCI interaction reproduced from d in black. (f) Plotting C-lobe CDI versus Ka,EFF under elevated Ca2+ unveils a well-resolved Langmuir relation. WT (green), I[0]A (red), and Q[1]A (blue).

Initially disappointing was the existence of only low-affinity binding between IQ and PCI modules (Fig. 6c, left cartoon; Fig. 6d, gray) under conditions of resting intracellular calcium3. By contrast, under elevated Ca2+, robust interaction between the same IQ/PCI FRET pair was observed, with Ka-PCI–IQ=4.35 × 10−5 Dfree units−1 (Fig. 6d, black). In fact, this Ca2+-dependent interaction accords well with a role in triggering CDI, and likely arises from a requirement for Ca2+/CaM to bind the PCI domain before appreciable IQ association occurs (Supplementary Note S13). Beyond mere binding, however, functionally relevant interaction would be decreased by the same IQ-domain mutations that reduced C-lobe CDI. In this regard, IQ peptides bearing I[0]A or Q[1]A substitutions actually demonstrated strong and graded reductions in affinity (Fig. 6e, respective red and blue symbols), coarsely matching observed deficits in C-lobe CDI (Fig. 6b). Figure 6c summarizes the results of these and other FRET-binding assays (Supplementary Note S14) performed for loci with the strongest effects on C-lobe CDI (Fig. 6b, coloured bars under dashed black threshold). With these data, quantitative iTL analysis could be undertaken, where the presumed CDI transition in question would be the γ2 transduction step in Fig. 2a, and the relevant form of equation 1 is specified in Supplementary Note S15. Remarkably, plotting C-lobe CDI (Fig. 6b) versus IQ/PCI-binding affinity (Fig. 6c) indeed resolves a Langmuir (Fig. 6f). Thus, C-lobe CDI likely requires a tripartite complex of IQ, PCI, and C-lobe Ca2+/CaM (Fig. 6a).

ApoCaM preassociation within the PCI domain

Having explored Ca2+/CaM, we turned to apoCaM interactions. Elsewhere42 we have shown that apoCaM preassociates with a surface that at least includes2,4,5,20 the IQ element. Furthermore, homology modelling42 of a related apoCaM/IQ structure for NaV channels44,45 suggests that the CaV1.3 IQ module interacts with the C-lobe of apoCaM. Will the N-lobe of apoCaM then bind the PCI domain (Fig. 7a)? If so, then our earlier PCI mutations could have weakened N-lobe apoCaM interaction, and potentially diminished CDI by favouring configuration E channels (Fig. 1a, incapable of inactivation). This effect would not have been apparent thus far, as we invariably overexpressed CaM. However, with only endogenous CaM present in Fig. 7e, CDI reflects the operation of a system that includes configuration E (left schematic), and the observed CDICaMendo is indeed strongly attenuated by mutations at many loci (rose and red bars).

Figure 7: Footprint of apoCaM preassociation with the PCI segment.
figure 7

(a) Channel cartoon depicting apoCaM preassociated with the CI region, with C-lobe engaging IQ domain, and N-lobe associated with PCI region. (b) Whole-cell currents for TVM→AAA mutant in the PCI segment (Ca2+ in red; Ba2+ in gray), with only endogenous CaM present. Horizontal scale bar, 100 ms; vertical scale bar, 0.2 nA Ca2+ current. (c) Overexpressing CaMWT rescues CDI for TVM→AAA mutation, suggesting that PCI region harbours an apoCaM preassociation locus. Format as in b. (d) 33-FRET binding curves show strong apoCaM binding to CI region. WT in black; TVM→AAA in red. Each symbol, mean±s.e.m. of ~7 cells. (e) Bar-graph summary of CDI with only endogenous CaM present (CDICaMendo), across alanine scan of PCI region. TVM→AAA (red) shows strongest effect, with rose colored bars also showing appreciable CDI reduction. Bars, mean±s.e.m. of ~5 cells. Left, schematic of relevant CaM subsystem. (f) Bar-graph summary of CDI rescue upon overexpressing CaMWT (CDICaMhi), for mutations showing significant loss of CDI (coloured bars in e), or chosen at random (hashed bars in e). Bars, mean±s.e.m. of ~5 cells. Corresponding subsystem of regulation on the left. (g) Bar-graph summary of Ka,EFF for apoCaM binding to CI region, with partners as sketched on the left. Data obtained for nearly all mutants with significant CDI reduction (coloured in e), and for some mutants chosen at random (hashed in e). Error bars, non-linear estimates of s.d. (h) iTL analysis confirms role of PCI as functionally relevant apoCaM site. Plotting CDICaMendo/ CDICaMhi (e,f) versus Ka,EFF for apoCaM/CI binding uncovers well-resolved Langmuir relation. TVM→AAA, red; WT, green. (i) Overlaying like data for IQ-domain analysis presented elsewhere42 (blue symbols here) displays remarkable agreement, consistent with the same apoCaM binding both PCI and IQ domains. Deep blue symbols, A[−4], I[0], F[+4] hotspots for apoCaM interaction with IQ element.

We tested for decreased preassociation as the basis of this effect, by checking whether CDI resurged upon strongly overexpressing CaMWT. This manoeuvre should act via mass action to eliminate CaM-less channels20, restrict channels to the subsystem on the left of Fig. 7f and restore CDI. For all loci demonstrating appreciable reduction of CDICaMendo (Fig. 7e, rose and red), CDI was measured under strongly overexpressed CaMWT (CDICaMhi), as summarized in Fig. 7f. As the baseline, we confirmed that elevating CaMWT hardly affected CDI of WT channels (compare WT, dashed green lines in Fig. 7e). The high apoCaM affinity of WT channels renders configuration E channels rare, even with only endogenous apoCaM present20. By contrast, the TVM mutant exhibits an impressive return of CDI upon elevating CaMWT (Fig. 7b), as do many other mutants (Fig. 7f and Supplementary Note S16). Critically, scrutiny of the underlying configurations (Fig. 7e, left) reveals that CDICaMendo=CDICaMhi × Fb, where Fb is the fraction of channels prebound to apoCaM with only endogenous CaM present. This relation holds true, even with a residual CDICaMhi shortfall compared with WT (for example, Fig. 7f, LGF). This is so, because a CDICaMhi deficit mirrors changes in the diamond subsystem of Fig. 7f (left), which are identically present in CDICaMendo measurements (Fig. 7e).

Thus aware, we tested whether apoCaM binding to the entire CI domain (spanning IQ and PCI modules) mirrors resurgent CDI (Fig. 7g, left cartoon; Supplementary Note S17). The WT pairing showed robust interaction with Ka,EFF=2.5 × 10−4 Dfree units−1 (Fig. 7d, black; Fig. 7g, green dashed line). By contrast, the TVM pairing, corresponding to strong resurgent CDI (Fig. 7b), exhibited far weaker affinity (Fig. 7d, red; Ka,EFF=0.13 × 10−4 Dfree units−1). Figure 7g tallies the graded decrease of Ka,EFF for these and other pairings (Supplementary Note S16).

Most rigorously, if PCI contacts indeed mediate apoCaM preassociation, then plotting CDICaMendo/CDICaMhi (=Fb) versus the Ka,EFF for apoCaM/CI interaction (Fig. 7g) should decorate a Langmuir relation (Supplementary Note S18). Indeed, just such a relation (Fig. 7h) is resolved (Fig. 7e–g), arguing that the N-lobe of apoCaM interfaces with corresponding PCI loci. Notably, this relation is identical to that reported elsewhere for IQ mutations on the same CI module42. Figure 7i explicitly overlays PCI and IQ data (in blue), and this striking resolution of a single Langmuir accords with one and the same apoCaM binding PCI and IQ domains.


These experiments fundamentally transform the prevailing molecular view of CaM regulation of Ca2+ channels. The field has long been dominated by an IQ-centric scheme2,5,7,9,10,25,28, wherein indwelling apoCaM begins preassociated with a C-terminal IQ domain, and remains bound to this element upon CaM interaction with Ca2+. Here our new proposal establishes substantial exchange of CaM to alternate effector loci (Fig. 8a). ApoCaM preassociates with an interface that includes, but is not limited to, the IQ domain (configuration A): the C-lobe binds the IQ (cyan circle) and the N-lobe binds the central midsection of the PCI (green box). Ca2+ binding to the N-lobe yields configuration IN, wherein this lobe binds the NSCaTE module on the channel N terminus (pink oval) to trigger the N-lobe CDI. Ensuing Ca2+ binding to the C-lobe induces configuration ICN, with C-lobe binding the proximal PCI midsection (green square) and IQ engagement25. If Ca2+ only binds the C-lobe, the system adopts configuration IC, corresponding to C-lobe CDI. Ultimately, Ca2+/CaM exchange to effector loci diminishes opening, perhaps via allosteric coupling of carboxy-tail conformation to a contiguous IVS6 segment implicated in activation46,47 and inactivation48. Only a single CaM is present22,23 (Supplementary Note S17).

Figure 8: New view of CaM regulatory configurations of CaV1.3 channels.
figure 8

(a) Molecular layout of configurations A, IC, IN and ICN for conceptual scheme in Fig. 1a. ApoCaM preassociates with CI region: C-lobe articulates IQ domain, and N-lobe engages the PCI segment. Once Ca2+ binds CaM, the N-lobe of Ca2+/CaM departs to NSCaTE on channel N terminus, eliciting N-lobe CDI (IN). Alternatively, the C-lobe of Ca2+/CaM migrates to PCI segment, recruiting IQ domain to tripartite complex (IC). Finally, ICN corresponds to channel that has undergone both N- and C-lobe CDI. (b) De novo model of CaV1.3 CI region docked to apoCaM (PCI region: green; IQ domain: blue). ApoCaM hotspots (Fig. 6e–g) in red. C-lobe of apoCaM contacts IQ, whereas N-lobe binds EF-hand region. (c) Left, atomic structure of NSCaTE bound to N-lobe of Ca2+/CaM (2LQC33). NSCaTE peptide in tan; and N-lobe Ca2+/CaM in cyan. Ca2+, yellow. N-lobe CDI hotspots on NSCaTE in red. Right, de novo model of tripartite IQ-PCI-Ca2+/CaM complex (PCI region, green; IQ domain, blue). C-lobe CDI hotspots in red for both PCI and IQ domains.

The structures of many of these configurations are presently unknown, but ab initio and homology modelling here confirms the plausibility of these configurations. Concerning the apoCaM/channel complex, Fig. 8b displays a homology model of the C-lobe complexed with the IQ domain42 (blue), based on an analogous atomic structure from NaV channels44,45. Key IQ-domain hotspots for apoCaM preassociation (red) are rationalized by this model42. To portray the N-lobe as shown in Fig. 8b, we utilized ab initio structural prediction of the CI domain with the Rosetta package49 (Supplementary Note S19), yielding a PCI domain (green) with two vestigial EF hands, and a protruding helix (‘preIQ’ subelement). The EF-hand module (EF) resembles the structure of a homologous segment of NaV channels50,51, and a helical segment has been resolved in atomic structures of analogous CaV1.2 segments36,37. Reassuringly, N-lobe apoCaM hotspots adorn the surface of this PCI model (red coloration), within the more C-terminal of the two EF hands. Accordingly, we appose the atomic structure of the N-lobe (1 CFD) to this segment of the PCI model, initially using a shape-complementarity docking algorithm52 (PatchDock), followed by refinement with docking protocols of Rosetta (Supplementary Note S20). Of note, the configuration of the N-lobe explains the outright enhancement of N-lobe CDI by PCI mutations in the region of putative N-lobe contact (compare Figs 5e and 7e, GKL through TLF). Weakening channel binding to the N-lobe (Fig. 2a, state 1) would, through connection to other states, increase state 5 occupancy, thereby boosting N-lobe CDI (Supplementary Note S21). By contrast, no N-lobe CDI enhancement was observed for IQ substitutions at the central isoleucine (I[0]) and downstream42, consistent with IQ binding the C-lobe of apoCaM.

Figure 8c displays a model of Ca2+/CaM complexed with the channel. The N-lobe bound to NSCaTE is an NMR structure33, and functional N-lobe CDI hotspots correspond well with intimate contact points (red). C-lobe CDI hotspots also adorn the surface of the upstream EF-hand region of an alternative ab initio model of the PCI (Fig. 8c, red; Supplementary Note S19). The IQ domain and atomic structure of the C-lobe of Ca2+/CaM (3BXL) were then computationally docked (Supplementary Note S22), yielding a rather canonical CaM/peptide complex where the channel contributes a surrogate lobe of CaM. Overall, this framework promises to set the table for future structural biology and structure–function work.

More broadly, this regulatory scheme may explain paradoxes and open horizons. First, it has been asked how Ca2+/CaM could ever leave the IQ domain, when the binding affinity between these elements is so high4,5,29,39,53 (for example, Ka/CaM–IQ=5.88 × 10−4 Dfree−1 units in Fig. 3c). The answer may arise from the competing binding affinity for the tripartite complex (Fig. 8a, ICN), which multivalent ligand binding theory54 would approximate as Ka/CaM–PCI–IQ~Ka/CaM–PCI × Ka/PCI-IQ × (local concentration of IQ)=(4.35 × 10−5 Dfree units−1) × (3.45 × 10−5 Dfree units−1) × (1.36 × 108 Dfree units)~0.2 Dfree units−1, a value far larger than Ka/CaM–IQ (Supplementary Note S23). Second, our scheme offers new interfaces targetable by native modulators and drug discovery. As L-type channel CDI influences cardiac arrhythmogenic potential15,18 and Ca2+ load in substantia nigral neurons prone to degeneration in Parkinson’s55, one could envisage a screen for selective modulators of N- or C-lobe CDI19. Third, our results offer a fine-grained roadmap for CaV1-2 splice/editing variants and channelopathies56. Indeed, we suspect that the design principles revealed here may generalize widely to other molecules modulated by CaM45,50,51,57.


Molecular biology

To simplify mutagenesis, the WT construct in this study was an engineered CaV1.3 construct α1DΔ1626, nearly identical to and derived from the native rat brain variant (α1D, AF3070009). Briefly, the α1DΔ1626 construct, as contained with mammalian expression plasmid pCDNA6 (Invitrogen), features introduction of a silent and unique KpnI site at a position corresponding to ~50 amino acid residues upstream of the C-terminal IQ domain (G1538T1539). As well, a unique BglII restriction site is present at a locus corresponding to ~450 amino acid residues upstream of the IQ domain. Finally, a unique XbaI and stop codon have been engineered to occur immediately after the IQ domain. These attributes accelerated construction of cDNAs encoding triple alanine mutations of α1DΔ1626. Point mutations of channel segments were made via QuikChange mutagenesis (Agilent) before PCR amplification and insertion into the full-length α1DΔ1626 channel construct via restriction sites BglII/KpnI, KpnI/XbaI or BglII/XbaI. Some triple alanine mutation constructs included a seven amino acid extension (SRGPAVRR) after residue 1626. For FRET two-hybrid constructs, fluorophore-tagged (all based on enhanced cyan fluorescent protein (ECFP) and enhanced yellow fluorescent protein (EYFP)) CaM constructs were made as described4. Other FRET constructs were made by replacing CaM with appropriate PCR-amplified segments, via unique NotI and XbaI sites flanking CaM4. YFP–CaMC (Supplementary Fig. S10) was YFP fused to the C-lobe of CaM (residues 78–149). To aid cloning, the YFP-tagged CI region was made with a 12 residue extension (SRGPYSIVSPKC) via NotI/XbaI sites as above. This linker did not alter apoCaM binding affinity versus WT YFP-tagged CI region (not shown). Throughout, all segments subject to PCR or QuikChange mutagenesis were verified in their entirety by sequencing.

Transfection of HEK293 cells

For whole-cell patch clamp experiments, HEK293 cells were cultured on 10-cm plates, and channels transiently transfected by a calcium phosphate method9. We applied 8 μg of complementary DNA encoding the desired channel α1-subunit, along with 8 μg of rat brain β2a (M80545) and 8 μg of rat brain α2δ (NM012919.2) subunits. We utilized the β2a auxiliary subunit to minimize voltage-dependent inactivation. For experiments involving CaM overexpression, we coexpressed 8 μg of rat CaMWT, CaM12 or CaM34, as described9. All of the above cDNA constructs were included within mammalian expression plasmids with a cytomegalovirus promoter. To boost expression, cDNA for simian virus 40T antigen (1–2 μg) was cotransfected. For FRET two-hybrid experiments, HEK293 cells were cultured on glass-bottom dishes and transfected with FuGENE 6 (Roche) before epifluorescence microscope imaging4. Electrophysiology/FRET experiments were performed at room temperature 1–3 days after transfection.

Whole-cell recording

Whole-cell recordings were obtained using an Axopatch 200 A amplifier (Axon Instruments). Electrodes were made from borosilicate glass capillaries (World Precision Instruments, MTW 150-F4) yielding 1–3 MΩ resistances, which were in turn compensated for series resistance by >70%. Currents were low-pass filtered at 2 kHz before digital acquisition at several times that frequency. A P/8 leak-subtraction protocol was used. The internal solution contained (in mM): CsMeSO3, 114; CsCl, 5; MgCl2, 1; MgATP, 4; HEPES (pH 7.4), 10; and BAPTA (1,2-bis(o-aminophenoxy)ethane- N,N,N′,N′-tetraacetic acid), 10; at 290 mOsm adjusted with glucose. The bath solution was (in mM): TEA-MeSO3, 102; HEPES (pH 7.4), 10; CaCl2 or BaCl2, 40; at 300 mOsm, adjusted with TEA-MeSO3.

FRET optical imaging

We conducted FRET two-hybrid experiments in HEK293 cells cultured on glass-bottom dishes, using an inverted fluorescence microscope as extensively described by our laboratory4. Experiments utilized a bath Tyrode’s solution containing either 2 mM Ca2+ for experiments probing apoCaM binding or 10 mM Ca2+ with 4 μM ionomycin (Sigma-Aldrich, MO) for elevated Ca2+ experiments. The 33-FRET efficiencies (EA) were computed as elaborated in our prior publications4. E-FRET efficiencies (ED), whose measurement methodology was developed and refined in other labortories41, could be determined from the same single-cell 33-FRET measurements using the following relationship, which expresses ED in terms of our own calibration metrics and standard measurements:

where G is a constant, defined as

SFRET, SYFP and SCFP correspond to fluorescent measurements from the same cell using FRET, YFP and CFP cubes, whose spectral properties have been detailed previously4. RD1 and RA are constants relating to the respective spectral properties of ECFP and EYFP; εCFP (440 nm)/εYFP(440 nm) approximates the ratio of molar extinction coefficients of ECFP and EYFP at 440 nm, respectively; and MA/MD is the ratio of optical gain factors and quantum yields pertaining to EYFP and ECFP, respectively. Detailed descriptions of these parameters and their determination appear in our prior publications4. For all FRET efficiencies, spurious FRET relating to unbound ECFP and EYFP moieties has been subtracted13. For 33-FRET, spurious FRET is linearly proportional to the concentration of CFP molecules, and the experimentally determined slope was obtained from cells coexpressing ECFP and EYFP fluorophores. Similarly, for E-FRET, the spurious FRET is linearly proportional to the concentration of EYFP molecules. The slope for this relationship AE−FRET can be obtained from:

The methods for FRET two-hybrid binding curves have been extensively described in previous publications3,4,41. Briefly, binding curves were determined by least-squared error minimization of data from multiple cells, utilizing a 1:1 binding model, with adjustment of parameters Kd,EFF and maximal FRET efficiency at saturating donor concentrations. For a small number of interactions involving mutations that strongly disrupted binding, the maximal FRET efficiency was set equal to that of the corresponding WT interaction and Kd,EFF varied to minimize errors. The s.d. error bounds on Kd,EFF estimates were determined by Jacobian error matrix analysis58.

Supplementary Note S24 characterizes our FRET two-hybrid constructs, specifying their precise sequence composition, and behaviour via western blot and confocal imaging analysis.

Molecular modelling

De novo structural prediction was performed using the Robetta online server49 ( as described in Supplementary Notes S19, S20–S22. We used web-based molecular docking programs, PatchDock52 ( and FireDock59 ( to obtain preliminary models for molecular docking. Such preliminary models were subsequently used as starting models for further structural modelling and refinement using a customized docking protocol of PyRosetta60. A homology model of the C-lobe of apoCaM bound to IQ domain was constructed as described elsewhere42. All molecular models and atomic structures were visualized and rendered using PyMOL v1.2r1. (DeLano Scientific, LLC).

Additional information

How to cite this article: Ben Johny, M. et al. Dynamic switching of calmodulin interactions underlies Ca2+ regulation of CaV1.3 channels. Nat. Commun. 4:1717 doi: 10.1038/ncomms2727 (2013).