Ca²⁺ channel nanodomains boost local Ca²⁺ amplitude

Significance

Local Ca²⁺ signals drive important events like synaptic transmission, neural plasticity, and cardiac contraction, yet the fundamental relationship between flux through a single channel and the Ca²⁺ amplitude within nanometers of the channel pore has eluded direct experimental measure. Here we reverse-engineer the problem by using Ca²⁺-dependent inactivation of channels as a nanometer-range biosensor of Ca²⁺ amplitude. We discovered an unexpected and dramatic boost in nanodomain Ca²⁺ amplitude, ten-fold higher than predicted on theoretical grounds. This boost in local Ca²⁺ signaling may act to maximize the biochemical information capacity of electrically\x{2011}active cells.

Abstract

Local Ca²⁺ signals through voltage-gated Ca²⁺ channels (Ca_Vs) drive synaptic transmission, neural plasticity, and cardiac contraction. Despite the importance of these events, the fundamental relationship between flux through a single Ca_V channel and the Ca²⁺ signaling concentration within nanometers of its pore has resisted empirical determination, owing to limitations in the spatial resolution and specificity of fluorescence-based Ca²⁺ measurements. Here, we exploited Ca²⁺-dependent inactivation of Ca_V channels as a nanometer-range Ca²⁺ indicator specific to active channels. We observed an unexpected and dramatic boost in nanodomain Ca²⁺ amplitude, ten-fold higher than predicted on theoretical grounds. Our results uncover a striking feature of Ca_V nanodomains, as diffusion-restricted environments that amplify small Ca²⁺ fluxes into enormous local Ca²⁺ concentrations. This Ca²⁺ tuning by the physical composition of the nanodomain may represent an energy-efficient means of local amplification that maximizes information signaling capacity, while minimizing global Ca²⁺ load.

Local Ca²⁺ signals increase the information capacity of signaling within a cell, by enabling short-range signals to operate independently of Ca²⁺ events elsewhere throughout the cell (1, 2). These local Ca²⁺ signals are the conduit that drives diverse activity-dependent events, such as synaptic transmission (3), neural plasticity (4–6), and cardiac excitation–contraction coupling (7). However, with regard to the Ca²⁺ amplitude within nanometers of individual Ca²⁺ channels, our knowledge is based predominantly on diffusion theory, which predicts ∼100 μM Ca²⁺ signals per picoampere flux at a distance of ∼10 nm from the pore (8–11). Although often quoted, this theory has been difficult to verify experimentally because diffusible Ca²⁺ indicators report space-averaged [Ca²⁺], and thus lack the spatial resolution needed for selective nanodomain reporting. Two recent efforts to overcome this limitation used voltage-gated Ca²⁺ channel (Ca_V)-tethered fluorescent Ca²⁺ indicators to isolate nanodomain signals (12, 13), but were met with unexpected difficulties: Ca_V-tethered indicators were unresponsive in the presence of high concentrations of intracellular 1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid (BAPTA), a Ca²⁺ buffer that preserves Ca²⁺ elevations within tens of nanometers of active channels but eliminates it elsewhere (8–10). A lack of response under these conditions would arise if the majority of tethered channels were silent, raising the concern that fluorescence changes observed under milder Ca²⁺ buffering (12, 13) may represent the response of indicators tethered to silent channels, which nonetheless eavesdrop on Ca²⁺ from active channels nearby. To overcome this fundamental limitation, we used a reverse strategy, whereby Ca²⁺/calmodulin–dependent inactivation (CDI) of channels themselves serves as a nanometer-range indicator of [Ca²⁺]; a sensitive ionic readout that is unaffected by the presence of silent channels, and exquisitely specific to [Ca²⁺] in the nanodomain of active channels.

Our strategy was to calibrate CDI with spatially uniform Ca²⁺ uncaging, and then use CDI to measure Ca_V-generated Ca²⁺ signals. We chose Ca_V1.3 because it displays prominent CDI, driven by Ca²⁺ binding to calmodulin (CaM), an indwelling subunit of the channel (14–16). To calibrate CDI, we generated Ca²⁺ elevations via intracellular Ca²⁺ uncaging, while monitoring CDI via the flux of Li⁺ ions, which do not bind to CaM or Ca²⁺ indicators (Fig. 1A). We engineered the selectivity filter of the channel to favor the flux of Li⁺ permeation in the face of Ca²⁺ uncaging (refs. 17, 18; SI Appendix, Fig. S1), and restricted Ca²⁺ binding to the C lobe of CaM₁₂ to isolate the simplest form of CDI (16).

Fig. 1.

Calibration of CDI. (A) CDI calibration schematic. Intracellular Ca²⁺ uncaging produces spatially uniform [Ca²⁺] elevations that drive CDI, as monitored via decay in the flux of Li⁺ ions. Use of a mutant CaM₁₂ restricts Ca²⁺ binding to the C lobe to isolate the simplest form of CDI. (B) Step depolarization alone (Top) produced noninactivating Li⁺ currents (Bottom, black trace). Uncaging (arrow, 30 ms after depolarizing step) produced [Ca²⁺] elevations (Middle, red trace) and CDI (Bottom, red trace). CDI is the fraction of current lost 800 ms after uncaging (Right). This metric approximates steady-state CDI. (C) Nonspecific Ca²⁺ uncaging effects determined with CDI eliminated via Ca²⁺-insensitive CaM₁₂₃₄. (D) [Ca²⁺] uncaging to ∼3 μM had minimal effects on Li⁺ current. Format as in B. (E) C lobe CDI plotted as a function of uncaged [Ca²⁺] (as in B). Mild nonspecific effects (e.g., D) are accounted for as in SI Appendix, Fig. S1. Data from 51 cells followed a binding relation with EC₅₀ = 2.2 ± 0.1 μM (dashed lines corresponding to half-maximal CDI) and Hill coefficient n = 1.5 ± 0.1.

Because photolysis is spatially uniform, the free Ca²⁺ concentration in the nanodomain will equal Ca²⁺ elsewhere within the cell, where it can be accurately determined with ratiometric indicators (19–21) (Fig. 1B, Middle row; Materials and Methods; and SI Appendix, Fig. S1). Before uncaging, step depolarization (Fig. 1B, Top) produced noninactivating Li⁺ currents (Fig. 1B, Bottom, black trace), consistent with an absence of voltage-dependent inactivation (22). Uncaging to low [Ca²⁺] evoked a small amount of inactivation (Fig. 1B, Left, red trace), and steps to higher [Ca²⁺] resulted in stronger CDI (Fig. 1B, Right). In parallel experiments, CDI was virtually eliminated with a Ca²⁺-insensitive CaM₁₂₃₄ (23, 24) (Fig. 1 C and D). This confirmed that uncaged Ca²⁺ had little direct effect on Li⁺ permeation, and enabled determination of CDI mediated entirely by Ca²⁺ binding to the C lobe of CaM (SI Appendix, Fig. S1). Pooled data from 51 cells followed a tight binding curve with half-maximal effective concentration, EC₅₀ = 2.2 ± 0.1 μM and Hill coefficient n = 1.5 ± 0.1 (Fig. 1E). This procedure was tantamount to the familiar live-cell calibration of a fluorescent Ca²⁺ indicator, with the distinction that because CDI is an ionic measure, it will be precisely specific to active (rather than silent) Ca_V channels.

We turned next to CDI mediated by local Ca²⁺ elevations generated by flux through Ca_V channels (Fig. 2A). A graded degree of CDI was observed with variations of external Ca²⁺ concentration, using Ca²⁺ as the only external charge carrier. CDI averaged ∼50% at 0.05 mM [Ca²⁺]_o (Fig. 2B). As before, we isolated Ca²⁺ binding to the C lobe of CaM. To be sure that CDI originated entirely from Ca²⁺ flux through the home channel, Ca²⁺ spread was restricted with 10 mM BAPTA, a Ca²⁺ chelator that eliminates all but the local Ca²⁺ within tens of nanometers of the channel (8–10). To standardize the voltage-dependent channel conformation, measurements of CDI were always performed with step depolarizations 30 mV positive to the voltage of half-activation, V_1/2 (SI Appendix, Fig. S2), thus counterbalancing changes in surface-charge screening at varying [Ca²⁺]_o.

Fig. 2.

Measuring Ca_V-generated local Ca²⁺ amplitude. (A) Schematic use of CDI to measure nanodomain Ca²⁺ amplitude. (B) Local Ca²⁺-mediated CDI data. Colored traces correspond to external solutions with Ca²⁺ as the only charge carrier. These contained 40 (red), 0.6 (orange), 0.15 (green), and 0.05 mM (blue) external [Ca²⁺] (Materials and Methods). Black trace corresponds to 40 mM Ba²⁺ data, a reference for voltage-dependent inactivation. Intracellular solutions contained 10 mM BAPTA. Traces show mean ± SEM of 5–22 cells for each condition. CDI is the fraction of current lost after 800 ms (as in Fig. 1B). To counterbalance surface charge screening, all step depolarizations were chosen 30 mV positive to V_1/2, the voltage of half-activation (SI Appendix, Fig. S2). (C) Single-channel data. Voltage ramps (Top) evoked Ca_V1.3 single-channel Ba²⁺ currents (gray traces). Patches containing a single channel were identified by currents entirely bounded by the closed- (zero line) and open-channel (dashed line) levels (SI Appendix, Fig. S2). Hundreds of sweeps from five patches were averaged to yield <i_Ba>, the average single-channel flux in 40 mM Ba²⁺ (solid black curve overlaid with bottom sweep). At the relevant step potential (V_1/2 + 30 mV), <i_Ba> = 0.27 pA (black circle). (D) Whole-cell conversions. An identical ramp protocol (compare with C) was used in the whole-cell configuration. Black trace (40 mM Ba²⁺ data) provided a reference whole-cell current (i_WC) representing N·<i_Ba>, where N is the number of channels. Data from a given cell shared a common N, determined from the ratio of whole-cell to single-channel Ba²⁺ current. N typically ranged from 1,000 to 5,000. Scaling by this value (i_WC/N) converted whole-cell Ba²⁺ data to single-channel <i_Ba> amplitude (black symbol; 0.27 pA at V_1/2 + 30 mV), and whole-cell Ca²⁺ data to single-channel <i_Ca> amplitude (colored symbols). CDI was eliminated via coexpression of Ca²⁺-insensitive CaM₁₂₃₄, and voltages were specified relative to V_1/2 (SI Appendix, Fig. S2). Traces show mean ± SEM of seven cells. (E) Average single-channel Ca²⁺ flux, <i_Ca>, at V_1/2 + 30 mV (D) depended on external Ca²⁺ according to a binding relation with EC₅₀ ∼2 mM and Hill coefficient ∼1. Error bars, when larger than symbol size, show SEM, where error estimates were propagated through each numerical step. Note log scales. (F) CDI (as in B) plotted as a function of <i_Ca> (as in E) for each condition. Error bars show SEM. Data followed a binding relation with EI₅₀ = 2.15 ± 0.43 × 10⁻³ pA and Hill coefficient n =1.5 ± 0.1.

Having described how CDI varies with external Ca²⁺, we still needed to determine how unitary flux depends on [Ca²⁺]_o to obtain the desired signal–flux relationship. Direct measurements of unitary flux are feasible at physiological [Ca²⁺]_o of 2 mM (25, 26), but not at 0.05 mM, where CDI was half-maximal; here channel flux would be far smaller than instrumentation noise, precluding identification of patches containing a single channel. We overcame this difficulty by performing whole-cell recordings to measure I_Ca = N · <i_Ca>, where N is the number of active channels on a cell and <i_Ca> reflects the time-averaged flux through a single stochastically gating channel. N was specified by measuring I_Ba = N · <i_Ba> in the same cell, while knowing <i_Ba> from parallel single-channel measurements. Representative records (Fig. 2 C and D) illustrate this approach. In the single-channel configuration, voltage ramps (Fig. 2C, Top) yielded well-resolved unitary Ba²⁺ currents (Fig. 2C, Bottom, gray sweeps), enabling identification of patches containing a single channel. Averaging many such records yielded <i_Ba> (solid black curve), the average flux through a single channel. Thus, at the relevant depolarization (30 mV positive to V_1/2), <i_Ba> = 0.27 pA (Fig. 2C, black circle). In subsequent whole-cell experiments, N was determined for a given cell from the ratio of whole-cell to single-channel Ba²⁺ current, both having been performed with 40 mM Ba²⁺ as the charge carrier. Scaling by N converted whole-cell Ba²⁺ data to single-channel <i_Ba> amplitude (Fig. 2D, black symbol), and whole-cell Ca²⁺ data to single-channel <i_Ca> amplitude (Fig. 2D, colored symbols), where CDI was eliminated via coexpression of Ca²⁺-insensitive CaM₁₂₃₄. The ratio of <i_Ca> to <i_Ba> with 40 mM external divalent cation was ∼1/3 (Fig. 2D, black vs. red curves), as expected from prior single-channel studies (26, 27). Reducing external Ca²⁺ yielded progressively smaller currents (Fig. 2D), and <i_Ca> depended on external Ca²⁺ according to a binding relation with EC₅₀ ∼2 mM (Fig. 2E), consistent with prior estimates (26, 27).

Knowing how CDI and <i_Ca> each vary with [Ca²⁺]_o (Fig. 2 B and E) enabled CDI to be plotted as a function of <i_Ca>. The data conform to a conventional binding curve, with 50% CDI at a half-maximal effective flux, EI₅₀ = 2.15 ± 0.43 × 10⁻³ pA and Hill coefficient n = 1.5 ± 0.1 (Fig. 2F). Thus, the dependence of CDI on unitary Ca²⁺ flux (Fig. 2F) has nearly equivalent shape to its dependence on cytosolic [Ca²⁺] (Fig. 1E), both having indistinguishable Hill coefficients. The plots differ mainly with regard to the physical units along the x axis. Although Fig. 1E plots CDI as a function of [Ca²⁺] (μM), Fig. 2F plots CDI as a function of <i_Ca> (in picoamperes). The scaling factor needed to optimally align all data points from our two experimental configurations provides a conversion factor, here termed the integrated gain, G = EC₅₀ / EI₅₀, which has the desired dimensionality for relating Ca²⁺ concentration to channel flux. Our estimate of G = 1.0 ± 0.2 mM/pA is nominally an order of magnitude larger than the widely quoted estimate of 100 μM/pA based on diffusion theory (detailed below). Repeating these analyses using the initial rate of CDI (rather than steady state) yielded nearly identical results (SI Appendix, Fig. S3). This suggested that nanodomain Ca²⁺ signaling may be considerably more efficient than previously appreciated, but how can this difference be explained?

One possible basis for the high G is that CaM could be located right at the mouth of the channel, at a radial position r ≅ 0. If this “close-CaM” hypothesis were true, CDI would be completely insensitive to Ca²⁺ buffering (explained below). We tested this idea by raising the concentration of intracellular exogenous Ca²⁺ buffer from 10 to 60 mM BAPTA and repeating the measurement of G (Fig. 3A and SI Appendix, Fig. S3). Contrary to prediction, G decreased by more than threefold, down to 0.28 ± 0.04 mM/pA, indicating that local [Ca²⁺] at the sensor is not only higher than expected, but also quite susceptible to buffering, inconsistent with the close-CaM hypothesis.

Fig. 3.

Intersection of data and theory. (A) Determination of buffer sensitivity. Traces (Top) show mean ± SEM data of 7–22 cells with 60 mM intracellular BAPTA; each color corresponds to a different [Ca²⁺]_o (as in Fig. 2B). Bottom plots CDI as a function of <i_Ca> for each condition (data points). Format and 10 BAPTA curve (gray) are replicated from Fig. 2F. The 60 BAPTA data conformed to a binding relation with EI₅₀ = 7.74 ± 0.68 × 10⁻³ pA and Hill coefficient n =1.5 ± 0.1. (B) Theoretical open-channel gain (G_O) profiles (Eq. 1, with parameters in SI Appendix, Table S2). Profiles for varying BAPTA concentrations (green, black, and gray curves), with 1× (dashed) vs. 0.1× (solid curves) diffusion, relative to free-aqueous diffusion. (C) Theoretical integrated gain (G) profiles (Eq. 2) in the presence of 10 and 60 mM BAPTA (black and gray curves) for 1× diffusion. Experimental measures of G in 10 and 60 mM BAPTA are shown in horizontal blue and red lines, respectively. Intersection of data with theory pinpoints r_CaM, the distance of CaM from the channel pore (blue and red arrows). The lack of correspondence between estimates of r_CaM in each condition indicates that data are inconsistent with 1× diffusion. (D) Theoretical 0.1× diffusion profiles are larger and more buffer sensitive (cartoon and profiles). These produce r_CaM estimates consistent with data (blue and red arrows overlap). (E) Theoretical 0.01× diffusion profiles overshoot the data (blue and red arrows misaligned). (F) Fitting relation; r_CaM was determined (see blue and red arrows in C–E) for relative diffusion values between 1× and 0.01× for 10 and 60 mM BAPTA (blue and red curves, corresponding to G = 1.0 ± 0.2 mM/pA and G = 0.28 ± 0.04 mM/pA, respectively). Thick and thin curves correspond to analysis with mean ± SEM data. Data are most consistent with ∼0.1× diffusion and r_CaM ∼7 nm (as in D).

To reconcile these observations and their implications for Ca²⁺ signaling, we turned to classical Neher–Stern diffusion theory (8, 9), which predicts that in the nanodomain of an open channel, the [Ca²⁺]/flux ratio (G_O, open-channel gain) depends on two factors, reflecting dispersion and buffering:

where r is distance from the channel pore, D_Ca is the diffusion coefficient of free Ca²⁺ ions, and F is Faraday’s constant. In the absence of buffering, dispersion alone would produce a 1/r spatial gradient, reflecting the dilution of Ca²⁺ as it diffuses into hemispherical annuli of growing volume (Eq. 1 and Fig. 3B, dashed green curve). The buffering factor incorporates a length constant (λ, Eq. 1), reflecting the interplay between diffusion and buffering. Ca²⁺ buffers can intercept Ca²⁺ ions, but only with finite kinetics, specified by k_B,on (buffer association rate), and B_T (total concentration of diffusible intracellular buffer). Thus, buffering has no efficacy at r = 0 where new ions enter the cytoplasm for the first time, but grows in influence as time from first entry and distance from the pore increase (Fig. 3B, dashed black and gray curves). Eq. 1 holds for virtually the entire period of a channel opening because compact nanodomain dimensions ensure that Ca²⁺ influx, diffusion, and buffering all reach steady state within microseconds of channel opening (SI Appendix, Fig. S4). Similarly, [Ca²⁺] drops to baseline levels (∼0 mM with BAPTA present) within microseconds once the channel shuts (10, 11, 16). Taking account of the duty cycle of channel opening and closing characterized by channel open probability (P_O), and the cooperativity of CDI reflected by Hill coefficient (n), the integrated gain (G, derived in SI Appendix, section III) is given by

where κ, a correction factor for the intermittency of channel opening, would be unity if P_O = 1, and greater than unity if P_O < 1. In our case, P_O = 0.43 was measured from single-channel data (SI Appendix, Fig. S2), n = 1.5 was obtained from the steepness of CDI-binding relations (Figs. 1E, 2F, and 3A), and κ was thus determined to be ∼1.3.

Reckoning with channel intermittency enabled direct comparison of theory to data. Thus, our experimental measurements of G (Fig. 3C, blue and red horizontal lines) are at odds with spatial profiles predicted by diffusion theory with conventional parameter values (Eqs. 1 and 2), which require CaM to be very close to the pore with 10 mM BAPTA but farther away with 60 mM BAPTA (Fig. 3C, blue and red arrows misaligned), a highly implausible scenario. The conflict would be resolved if the diffusion coefficient (D_Ca) were reduced 10-fold below that of free aqueous diffusion. Then, the amplitude of nanodomain [Ca²⁺] would be increased (due to Ca²⁺ backlog), and the radial falloff of Ca²⁺ would be steepened (due to shorter λ). Ca²⁺ diffusion theory could then account for our seemingly contradictory findings of a boosted G in 10 BAPTA that drops significantly in 60 BAPTA, all with CaM at a fixed position relative to the pore (Fig. 3D, blue and red arrows overlap). Further reduction in D_Ca, to a level 100-fold below free aqueous diffusion causes theory to overshoot the data (Fig. 3E, blue and red arrows again discrepant). Accordingly, we performed a systematic analysis to determine how putative CaM position (r_CaM) would depend on variation in D_Ca (Fig. 3F). Intersection of plots for data obtained with 10 and 60 mM BAPTA puts narrow constraints on both Ca²⁺ diffusion (∼10-fold below free aqueous value), and CaM position (∼7 nm from the channel pore), as depicted in Fig. 3D. With this empirical estimate of D_Ca in hand, profiles of the open-channel gain (G_O, Eq. 1) were determined to be remarkably larger and more buffer sensitive than those based on aqueous diffusion (Fig. 3B, solid vs. dashed curves).

We asked whether the unexpectedly high nanodomain [Ca²⁺] under 10 mM BAPTA could arise from crosstalk among neighboring channels, which can occur with intracellular EGTA (due to large λ; Eq. 1) (8–10) or if Ca²⁺ buffers are depleted (10, 12). Indeed, these factors may explain why Ca_V-tethered fluorescent Ca²⁺ indicators are unresponsive in the presence of BAPTA (indicating a majority of silent channels) yet respond in a subset of cells loaded with EGTA [where Ca²⁺ can spread from active to neighboring silent channels; (12); see also ref. 13]. In the present study, high-buffer/low-flux conditions precluded buffer depletion (SI Appendix, Table S2 and Fig. S4), and the length constant (λ) imposed by 10 mM BAPTA is sufficiently short that Ca²⁺ spillover would be insignificant even if channels were packed at maximal density (SI Appendix, Fig. S5). These analyses support that our estimates are bona fide measures of G corresponding to a single active channel.

With regard to restricted nanodomain diffusion, we characterized the degree to which uncertainties in k_B,on (Eq. 1) could alter our conclusions. We found that a 10-fold change in k_B,on would only alter estimates of D_Ca by twofold, and that a substantial degree of diffusion restriction was absolutely necessary to explain our data (SI Appendix, Fig. S6). To estimate how far the region of restricted diffusion extends, we modeled cases in which diffusion was slowed only within the confines of the nanodomain (radius r_nano), but not outside it (Fig. 4). Remarkably, the same 10-fold reduction in diffusion would recapitulate our data even if it were isolated to a hemisphere of r_nano ∼15 nm (SI Appendix, Fig. S7). Smaller regions required further reductions in D_Ca to recapitulate data. Finally, we performed additional modeling to determine whether our estimates of D_Ca and r_CaM could be altered by volume exclusion (e.g., by nearby organelles or cytosolic Ca_V domains) (28–30), or by electrodiffusion (Ca²⁺ attraction toward electronegative phospholipids) (30, 31). We determined that these factors might alter estimates of r_CaM, but could only accentuate the required reduction in D_Ca (SI Appendix, Figs. S8 and S9).

Fig. 4.

Potential modes of physiological tuning. (A) Model of localized diffusion restriction extending over radius r_nano from the pore, with normal diffusion at a distance. (B) Resulting G_O profiles. Green profiles correspond to homogenous 1× and 0.1× profiles (replicated from Fig. 3B). Blue profiles correspond to conditions where the region of 0.1× diffusion is spatially isolated to r_nano = 20 or 35 nm (SI Appendix, Fig. S7). Local Ca²⁺ gain can be amplified by increasing the degree of diffusion restriction near the channel (vertical arrow). The spatial extent of amplification could be adjusted by modulating the radius of diffusion restriction (horizontal arrow).

In summary, using a method that distinguishes between active and silent channels, we found that the gain of Ca²⁺ signaling via Ca_V channels is greatly amplified by the diffusional properties of Ca²⁺ within tens of nanometers of the pore. To maximize experimental precision, our experiments were performed with Ca_V1.3 channels in HEK cells, but a similar approach could be extended to other Ca²⁺-permeant channels in various cellular contexts. Indeed, our use of CDI to measure nanodomain Ca²⁺ amplitude mirrors the prior use of K_Ca channels (32, 33) and synaptic transmission (19–21) as biosensors for presynaptic Ca²⁺ amplitude. Given these measures, canonical estimates of D_Ca have typically been required to infer distance from Ca_V to sensor. Here, we expand upon the basic biosensor approach to enable direct estimates of both D_Ca and r_sensor in the same preparation (Fig. 3F). In contrast to canonical parameters, our data indicate that Ca_V channels can amplify tiny Ca²⁺ fluxes into enormous Ca²⁺ concentrations, on the order of ∼1 mM per pA (Fig. 3B, solid green curve).

Our findings may have implications in diverse physiological contexts. With regard to biochemical signaling cascades, tiny ∼fA Ca²⁺ fluxes, generally believed not to partake in local signaling, could produce ∼μM Ca²⁺ signals sufficient to signal locally to molecules like CaM or CaMKII (34). Nanodomain boosting may similarly play a role in the contraction of ventricular myocytes, where controversy exists regarding whether activation of ryanodine receptors can be triggered by the opening of a single Ca_V channel [high-fidelity scenario supported by early experimental work (35–38)], or if the simultaneous opening of ∼10 channels is necessary [as suggested by recent models that take account of the small ∼fA flux through a single Ca_V channel at the +50 mV ventricular plateau potential (39–41)]. Our findings may reconcile this low-flux high-fidelity paradox given that a single Ca_V channel would, by virtue of nanodomain Ca²⁺ boosting, have the potency of ∼10 channels in the absence of this amplification. Similar implications may apply to presynaptic physiology, where the Ca²⁺ dependence of transmitter release has been experimentally determined (20, 21, 42, 43), and the sensitivity to exogenous buffers (EGTA and BAPTA) has been characterized in several preparations (44, 45). Whereas these findings are typically interpreted in the context of classical diffusion parameters, our findings may bear upon inferences of Ca_V–vesicle distance and stoichiometry.

With regard to mechanism, nanodomain Ca²⁺ boosting is most likely mediated by physical barriers to diffusion, termed tortuosity (28, 29), that slow the passage of free Ca²⁺ ions. Other potential means of diffusion reduction (e.g., fixed buffers), which limit movement by binding/immobilization of Ca²⁺ ions could only lower the concentration of free Ca²⁺ in the nanodomain, and thus are unlikely to play a role in our findings (SI Appendix, Fig. S4). Of note, the 10-fold reduction in diffusion of unbound Ca²⁺ ions implicated here by local reporting far exceeds the twofold attenuation previously advanced for bulk cytosol (46–48). On this view, the domain within close range of the channel is a remarkably diffusion-restricted environment, where a Ca²⁺ ion would collide with an obstacle in nine out of ten Brownian movements. This energy-efficient means of local amplification seems ideally suited to address a universal challenge of Ca²⁺ homeostasis: the toxicity of global Ca²⁺ overload, and energetic cost of Ca²⁺ extrusion. By locally amplifying Ca²⁺ signals produced by diminutive fluxes, nanodomains likely facilitate the maintenance of numerous independent Ca²⁺ sources, maximizing information signaling capacity, and minimizing global Ca²⁺ load.

It remains to be determined exactly how far the region of amplification extends and what the precise nature of the diffusion barriers might be. One possibility is that Ca_V cytosolic domains may physically impede diffusion over a ∼15 nm radius (SI Appendix, Fig. S7 and Fig. 4), as would be the case if Ca²⁺ ions passed through narrow Ca_V cytosolic crevices or portals (13, 49). Such a self-sufficient design would be robust to cellular context, and spatially isolated to the indwelling (50) and locally enriched pool (51) of CaM molecules mediating CDI and implicated in signaling to nuclear transcription factors (4, 5). Alternatively, the region of amplification could extend beyond the Ca_V itself, if augmented by diffusion-impeding assemblies of associated components. In this regard, it is interesting to note that Ca_V subtypes interact with a tremendous variety of proteins in different cellular contexts (52, 53). Indeed, the very same molecules that aggregate near Ca_V channels to convey downstream signals could also act to restrict diffusion, thereby increasing the gain between Ca²⁺ flux and biochemical response. It is similarly feasible that nanodomain boosting could be physiologically tuned by altering the density of associated components (Fig. 4B, vertical arrow, gain amplitude), or the radius over which associated components extend (Fig. 4B, horizontal arrow, gain radius). Such gain modulation would not only alter Ca²⁺ amplitude, but could also sharpen the Ca²⁺ profile, either by boosting Ca²⁺ in a restricted locale (Fig. 4) or by enhancing the efficacy of endogenous buffers like parvalbumin (54) (Fig. 3B). Thus, beyond amplification, diffusion-restricted nanodomains could play a role in curtailing crosstalk between Ca²⁺ sources, much like dendritic spines prevent Ca²⁺ spillover between adjacent postsynaptic sites (55).

Materials and Methods

Detailed methods can be found in SI Appendix. In brief, Rat α_1D (14) was engineered (E716A) to enhance permeation of Li⁺ over Ca²⁺ (17, 18), and expressed in HEK293 cells as described (23). This mutant was used in all experiments. Whole-cell and single-channel recordings were as described (16). Ca²⁺ uncaging (Fig. 1), with a Cairn UV flashlamp, was performed 3–6 min after break-in; internal solutions contained Ca²⁺-loaded DMNP-EDTA (1–4 mM). Uncaging transients were minimized with citrate (1–50 mM), zero Mg²⁺, and zero ATP. Internal solutions were supplemented with phosphatidylinositol 4,5-bisphosphate and okadaic acid to stabilize Ca_V currents in the absence of Mg²⁺/ATP. To enable ratiometric monitoring of [Ca²⁺], internal solutions contained a single-wavelength Ca²⁺ indicator (Fluo4FF) and red-shifted dye (Alexa568).