Nanodomain Ca²⁺ of Ca²⁺ channels detected by a tethered genetically encoded Ca²⁺ sensor

Abstract

Coupling of excitation to secretion, contraction, and transcription often relies upon Ca²⁺ computations within the nanodomain—a conceptual region extending tens of nanometers from the cytoplasmic mouth of Ca²⁺ channels. Theory predicts that nanodomain Ca²⁺ signals differ vastly from the slow submicromolar signals routinely observed in bulk cytoplasm. However, direct visualization of nanodomain Ca²⁺ far exceeds optical resolution of spatially distributed Ca²⁺ indicators. Here we couple an optical genetically encoded Ca²⁺ indicator (TN-XL) to the carboxyl tail of Ca_V2.2 Ca²⁺ channels, enabling nearfield imaging of the nanodomain. Under TIRF microscopy, we detect Ca²⁺ responses indicative of large-amplitude pulses. Single-channel electrophysiology reveals a corresponding Ca²⁺ influx of only 0.085 pA, and FRET measurements estimate TN-XL distance to the cytoplasmic mouth at ~55 Å. Altogether, these findings raise the possibility that Ca²⁺ exits the channel through the analog of molecular portals, mirroring the crystallographic images of side windows in voltage-gated K channels.

INTRODUCTION

Ca²⁺ signals extend pervasively across the intracellular expanse of most cells, yet these signals exhibit remarkable specificity in activating appropriate Ca²⁺-modulated targets. Understanding how this selectivity arises remains a foremost question in Ca²⁺ biology^1,2. One strategy is to colocalize at molecular dimensions Ca²⁺ sources and targets. This co-localization is crucial to the local signaling of Ca²⁺ channels to nearby Ca²⁺-regulated ion channels and enzymes^3–5, to neurotransmitter release⁶, as well as to excitation-contraction⁷ and excitation-transcription coupling^8,9. Intriguingly, colocalized sensors do not always respond preferentially to local Ca²⁺ sources, but can require the far weaker input of remote Ca²⁺ sources acting through a global Ca²⁺ selectivity paradigm¹⁰. Critical to signaling near Ca²⁺ sources is the nature of Ca²⁺ signals within a few tens of nanometers of a Ca²⁺ channel (nanodomain¹¹). Theoretical calculations^12–16 postulate that nanodomain signals comprise Ca²⁺ pulses of equal and enormous amplitude (~100 μM), with each pulse synchronized to the millisecond stochastic openings of channels^10,12. Beyond the nanodomain, [Ca²⁺] rapidly dissipates with diffusion. Of note, predicted nanodomain [Ca²⁺] amplitudes vary considerably with assumed parameters¹⁷, and elegant experimental estimates of local Ca²⁺ signals^18,19 only indirectly probe the nanodomain itself. Thus, the critical magnitude of nanodomain Ca²⁺ pulses (Ca_spike) has eluded explicit empirical comment. This uncertainty bears on key issues, including how spatial Ca²⁺ decoding occurs within the nanodomain^10,20–22, and the Ca²⁺ channel number needed to trigger neurotransmitter vesicles^23–25.

Why have nanodomain Ca²⁺ signals remained elusive? Though fluorescent Ca²⁺-sensitive chemical dyes has revealed much²⁶, visualization of nanodomain signals via freely diffusible dye molecules far exceeds the resolution limit of far-field microscopic approaches²⁷. In this regime, fluorescence from dye molecules outside the nanodomain will overshadow that from within²⁸. Even when imaged with total internal reflection fluorescence (TIRF) microscopy^29–31, which illuminates a restricted TIRF volume within ~150 nm of the glass/cell membrane interface, freely diffusible dyes will still report Ca²⁺ from a region several-fold larger than a Ca²⁺ channel³².

One approach is to affix a Ca²⁺ indicator within the nanodomain, so as to support a form of near-field imaging³³. The Ca²⁺-sensitive luminescent protein aequorin represents a traditional candidate³⁴. Recombinant aequorins have long been targeted to various subcellular compartments by genetic means^35–37, but the limited-amplitude and consumptive nature of aequorin would challenge deployment in a channel nanodomain³⁴. Recently, promising results have been obtained using a biarsenical chemical fluorescent Ca²⁺ indicator (Calcium Green FlAsH, CaGF) targeted to a tetracysteine tag implanted on Ca_V1.2 Ca²⁺ channels³⁸. CaGF features Ca²⁺ affinity and kinetics (K_d ~55 μM; off rate, 2 ms⁻¹) well matched to the purported amplitudes and lifetimes of nanodomain Ca²⁺ pulses. Indeed, CaGF was potentially responsive to nanodomain Ca²⁺ fluctuations, but the observed fluorescence signals were highly heterogeneous, to an extent that precluded quantitative Ca²⁺ estimation. This heterogeneity was attributable to the diminutive and variable open probability of Ca_V1.2 channels (~0.1–0.3%), despite constitutive pharmacological manipulation by a channel opener. Indeed, the possibility of a large fraction of electrically silent Ca_V1.2 channels fits with their unusually large ratio of gating charge to ionic current³⁹. Also, though the authors were careful to control for non-specific labeling of other cysteine-rich sites in the cell by CaGF, such indiscriminate attachment would remain a generic concern for approaches involving chemical targeting of sensors. A third approach would be to employ genetically encoded Ca²⁺ indicators (GECIs) as near-field sensors. GECIs allow genetic fusion of sensor to a Ca²⁺ source, offering the greatest potential selectivity in positioning sensors. Among the most popular GECIs are CFP/YFP-FRET-based molecules such as cameleons⁴⁰, and GFP-intensity-based sensors like GCaMPs⁴¹. All these employ calmodulin (CaM) as their Ca²⁺ sensor, and convert binding of the sensor’s own Ca²⁺/CaM with a target peptide into altered optical readouts. When targeted to plasmalemma, cameleons often lose Ca²⁺ sensitivity⁴², potentially via target peptide binding to elevated plasmalemmal CaM⁴³.

Accordingly, we here explore nearfield imaging of nanodomain Ca²⁺ using TN-XL, a GECI⁴⁴ based on the Ca²⁺ sensor troponin C. TN-XL exploits Ca²⁺-driven conformational changes without a target peptide, thereby naturally resisting endogenous CaM interference⁴². Our specific approach is to fuse TN-XL to the C-terminus of the principle subunit of Ca_V2.2 Ca²⁺ channels, chosen for high open probability⁴⁵, relative absence of silent channels⁴⁶, and abundant plasmalemmal expression⁴⁷. To further improve the TN-XL signal-to-noise ratio at the surface membrane, we utilize TIRF microscopy to selectively illuminate only those Ca_V2.2/TN-XL channels near the surface, thereby attenuating background signal from imperfectly targeted intracellular channels. Finally, to account for the high Ca²⁺ affinity and slow kinetics of GECIs⁴⁴ (off times 100–900 ms for TN-XL) in the face of millisecond nanodomain Ca²⁺ pulses, we quantify the kinetics of TN-XL readouts, which our previous work suggests can distinguish between differing Ca_spike amplitudes⁴⁸. These strategies allow us to resolve nanodomain Ca²⁺ activity of Ca_V2.2 channels undergoing native gating, without pharmacological manipulation.

RESULTS

Functional Ca_V2.2/TN-XL fusions at the surface membrane

Our strategy for probing nanodomain Ca²⁺ signals (Fig. 1) requires maintained Ca²⁺ responsiveness of genetically encoded sensors situated at the plasmalemma, a foremost challenge. Accordingly, we examined whether troponin-based TN-XL could respond to Ca²⁺ when targeted to this environment. As a baseline, Fig. 2a displays the properties of the well-studied, cytoplasmic form of TN-XL⁴⁴. Confocal images of an exemplar HEK293 cell, acquired under CFP and FRET imaging modes, reveal the expected pan-cytoplasmic distribution. An epifluorescence CFP image of another cell also demonstrates this generalized expression, and switching to TIRF mode indicates no preferential TN-XL accumulation within the plasmalemmal footprint of this same cell. To gauge TN-XL responsiveness, we measured under TIRF imaging the single-cell CFP (S_C, respective excitation and emission wavelengths at 442 and 485 nm) and FRET fluorescence signals (S_F, excitation and emission at 442 and 545 nm), where the ratio of these signals (R_F/C = S_F/S_C) furnishes the customary sensor readout of Ca²⁺ (Fig. 2b). In particular, the ratio R_F/C was first determined in resting cells with cytoplasmic Ca²⁺ concentration (≤ 0.1 μM) well below the K_d of the TN-XL²⁴, thus yielding R_min. To subsequently determine R_max, Ca_V2.2 channels were co-expressed as conduits of Ca²⁺ entry under whole-cell voltage clamp. To facilitate elevation of intracellular Ca²⁺, minimal internal Ca²⁺ buffering (1 mM EGTA) with elevated 10 mM extracellular Ca²⁺ was employed, such that repetitive step depolarizations (30 mV × 500 ms pulses every 15 s) readily saturated R_F/C at R_max. Thus, the ratio of R_max and R_min reports sensor dynamic range.

Figure 1. Approach to resolving channel nanodomain Ca²⁺ signals.

a, Conventional wide-field imaging using cytosolic chemical fluorescent dyes cannot resolve channel nanodomain Ca²⁺ signals. Blue shading denotes region of fluorescence excitation, which extends throughout the cell under wide-field imaging. b, Design of the genetically encoded Ca²⁺ indicator TN-XL fused to the carboxy tail of the α_1B subunit of a Ca_V2.2 channel, yielding Ca_V2.2/TN-XL. For orientation, structure-function elements involved in calmodulin regulation are denoted on carboxy terminus²¹: EF, EF-hand region; IQ, IQ-domain for apoCaM binding. CFP denotes enhanced CFP. YFP denotes circularly permuted Citrine. c, Ca_V2.2/TN-XL constructs act as a ‘near-field’ sensor of nanodomain Ca²⁺. TIRF imaging evanescent wave illuminates only Ca_V2.2/TN-XL channels within ~150 nm from the glass/cell membrane interface, as indicated by the blue-shaded region. This mode of excitation potentially excludes intracellular channels from consideration. When laser illumination angle Θ exceeds a critical angle Θ_C, TIRF illumination occurs.

Figure 2. Ca_V2.2/TN-XL fusion construct preserves function of sensor.

a, Behavior of free TN-XL. Left to right: Confocal (CFP filter), confocal (FRET filter), epi-fluorescence (CFP filter), and TIRF (CFP filter) images of HEK293 cells expressing cytoplasmic TN-XL. White scale bar indicates 10 μM. Bar at far left pertains to all confocal images. Bar at middle right pertains to all epifluorescence and TIRF images. b, TN-XL ratio (R_F/C=S_F/S_C) measured in resting cells (R_min) (gray bars) and in cells under high Ca²⁺ (R_max) (black bars). Data shown as mean ± sem with number of cells in parentheses. c and d, Behavior of TN-XL-Ras (membrane targeted TN-XL). Format as in a and b, respectively. e and f, Behavior of Ca_V2.2/TN-XL (N-type channel fused to TN-XL). Format as in a and b, respectively.

By contrast, for TN-XL fused to the farnesylation motif of Ras (TN-XL-Ras), sensors are sharply localized to the surface membrane in confocal and epifluorescence images (Fig. 2c), and there is marked enrichment within the TIRF footprint. Importantly, the Ca²⁺ dynamic range of TN-XL-Ras is nearly identical (Fig. 2d), despite residence within the plasmalemmal context. Yet more relevant are results for the Ca_V2.2/TN-XL fusions (Fig. 2e). Here, constructs appear only partially localized to the surface membrane, consistent with typically imperfect trafficking of recombinant Ca²⁺ channels to plasmalemma⁴⁹. Still, there is substantial focalization of Ca_V2.2/TN-XL in the TIRF footprint, suggesting sufficient trafficking for nanodomain experiments (Fig. 1c). Critically, sensor dynamic range is spared when juxtaposed near Ca_V2.2 channels (Fig. 2f).

A second concern is the preservation of Ca_V2.2 channel function upon fusion to TN-XL. Reassuringly, Ca²⁺ currents appeared nearly identical in Ca_V2.2/TN-XL (Fig. 3a, black) and Ca_V2.2 (gray) channels. In fact, fusing TN-XL to Ca_V2.2 enhances opening, as seen from hyperpolarization of tail-activation curves determined with 10 mM extracellular Ca²⁺ as charge carrier (Fig. 3b). Even at the single-channel level, Ca_V2.2/TN-XL fusions exhibited robust electrophysiological function, as illustrated by the exemplar single-channel current trace shown at high-gain magnification (Fig. 3c, noisy trace). We used 90 mM Ba²⁺ as a charge carrier to enhance resolution of elementary events, and evoked currents via ramp depolarization to readily resolve the open-channel conduction profile (downwardly convex gray relation). Multiple single-channel sweeps (Fig. 3d), displayed at lower-gain magnification, confirm the overall reproducibility and excellent resolution of elementary events under this protocol. Accordingly, numerous sweeps from multiple patches specify a highly reliable ensemble average current (Fig. 3c, red curve), closely similar to that of Ca_V2.2 channels without sensor fusion⁴⁵. Taking a ratio of the ensemble average current with the open-channel conduction profile (Fig. 3c, red and convex gray curves) yields a plot of open probability P_O versus voltage V (Fig. 3e, gray relation). This plot matches impressively with whole-cell activation (reproduced as open circles), after shifting along the voltage axis to account for contrasting surface-charge effects of single-channel versus whole-cell solutions. As with studies of Ca_V2.2⁴⁵, the Ca_V2.2/TN-XL fusion exhibits a substantial P_O of 0.69 at 30 mV (Fig. 3e, red arrows), a feature advantageous for probing nanodomain Ca²⁺.

Figure 3. Ca_V2.2/TN-XL fusion construct preserves function of channel.

a, Exemplar whole-cell currents of HEK293 cells expressing Ca_V2.2 channels (gray) or Ca_V2.2/TN-XL channels (black), using 10 mM Ca²⁺ as charge carrier, and 10 mM EGTA intracellular Ca²⁺ buffering. b, Whole-cell tail-activation curves of Ca_V2.2 (gray) and Ca_V2.2/TN-XL (black), obtained from records as in a. Data shown as as mean ± sem with number of cells in parentheses. c, Single-channel analysis of Ca_V2.2/TN-XL. Top, On-cell ramp-voltage waveform delivered every 5–15 s. Bottom, Exemplar single-channel sweep shown at high-gain magnification (100–200/patch). 90 mM Ba²⁺ as charge carrier; 2 kHz lowpass filtering. Leak-subtracted traces were averaged, yielding red I–V curve. Unitary current relation (convex gray curve) was fitted to the open-channel current level using GHK equation¹⁰. d, Multiple exemplar single-channel sweeps from another patch, shown at lower-gain magnification to impart a sense of the overall reproducibility and excellent resolution of elementary events. Format as in c. e, Single-channel P_O–V relation (gray, with fit in black, average of n = 4 patches), determined by dividing I–V curve (c, red trace) by unitary current relation (c, gray GHK fit). P_O–V curve was calibrated in voltage by aligning the single-channel activation curve with the whole-cell curve reproduced from b (open circles). From calibrated P_O–V curve, the P_o at 30 mV is 0.69. f, Immunoblots of HEK293 cell lysates probed with anti-GFP antibody; (left) Ca_V2.2/TN-XL at ~320 kD; (middle) TN-XL-Ras at ~71 kD; (right) control (untransfected) cells for reference. g, Lack of lower molecular-weight species in leftmost lane of panel f excludes proteolysis of sensor from channel.

A final prerequisite is that Ca_V2.2/TN-XL fusions resist proteolysis. Accordingly, we utilized anti-GFP antibody to perform western immunoblots from cells expressing this construct (Fig. 3f, left lane). The other lanes correspond to cells expressing TN-XL-Ras (~71 kD), as well as untransfected cells (no signal). The absence of signal beneath the full-length ~320 kD band (Fig. 3f, left lane) substantiates maintained TN-XL fusion to channels (Fig. 3g).

Calibration of TN-XL fused to Ca_V2.2 channels

We next addressed more precise sensor calibration, to facilitate quantitative inferences about nanodomain Ca²⁺. Both kinetic and steady-state characterization would be ideal, given the millisecond kinetics of Ca²⁺ pulses within channel nanodomains, coupled with the comparatively slow response of free TN-XL (off rates of 1/150 to 1/850 1/ms⁴⁴). Accordingly, we devised a state-mechanism approximation of sensor performance, a ‘forward transform’ that maps rapid Ca²⁺ transient inputs onto slower sensor outputs, as previously established for an older variant of TN-XL (TN-L15) freely expressed in myocytes⁴⁸. Here, we pursued an analogous approach, but attuned to Ca_V2.2/TN-XL constructs resident within the TIRF volume and driven by high-amplitude dynamic Ca²⁺ inputs pertinent to channel nanodomains. To produce such inputs, we activated ultra-large Ca_V2.2 Ca²⁺ currents (Fig. 4a) in HEK293 cells featuring minimal exogenous intracellular Ca²⁺ buffering (1 mM EGTA); this configuration intentionally produced enormous spatial gradients of intracellular Ca²⁺ (Fig. 4b inset), far exceeding that usually encountered. To estimate these gradients, we simultaneously monitored aggregate Ca²⁺ concentration (Fig. 4b, noisy dark trace), as deduced by imaging the entire cell for signals emanating from 10 μM of the rapid chemical fluorescent Ca²⁺ indicator Fluo 4FF. The measured Ca²⁺ influx (Fig. 4a) and aggregate [Ca²⁺] signals were then used to explicitly constrain a radially symmetric Ca²⁺ diffusion mechanism, whose estimates of Ca²⁺ concentration in differing cellular regions are shown as red traces in Fig. 4b (Supplementary Figure S1 and Supplementary Methods). The estimated aggregate [Ca²⁺] concentration (Fig. 4b, lower red trace) closely approximates its measured analog (dark noisy trace), and the estimated [Ca²⁺] at the surface membrane (upper red trace) peaks at ~80 μM and later converges to the aggregate [Ca²⁺] (~20–25 μM). This estimate of submembranous [Ca²⁺] was reproducible in multiple cells exhibiting like Ca²⁺ influx, and was thereby taken as the relevant Ca²⁺ input in parallel experiments performed with Ca_V2.2/TN-XL fusion constructs under TIRF imaging. The green trace in Fig. 4c displays the corresponding sensor output from GECIs in the submembranous TIRF volume, averaged from multiple cells exhibiting Ca²⁺ influx matching that in Fig. 4a. The steady-state plateau of the response (Fig. 4c, gray dashed line at R_F/C = 1.78), at a near steady Ca²⁺ concentration of ~20–25 μM, enabled us to calculate ~17.3 μM as the half-response point of the steady-state response curve for Ca_V2.2/TN-XL constructs in the TIRF volume (Supplementary Methods). This differs from the 2.5 μM half response point for free TN-XL⁴⁴, consistent with prior observations that sensor performance can change with cellular/molecular environment^38,42. Accordingly, steady-state response data for free TN-XL⁴⁴ were shifted along the [Ca²⁺] axis by ×16.3/2.5 to provide an appropriate steady-state profile for Ca_V2.2/TN-XL constructs (Fig. 4d, green symbols). In all, these steady-state data and the dynamic sensor response (Fig. 4c, green trace) to a specified submembranous Ca²⁺ input (Fig. 4b, upper red trace) furnished the constraints required to deduce a forward transform for Ca_V2.2/TN-XL.

Figure 4. Calibration of Ca_V2.2/TN-XL fusion construct.

a, High-amplitude Ca²⁺ influx into HEK293 cell expressing Ca_V2.2 channels, observed with 10 mM external Ca²⁺ and minimal internal Ca²⁺ buffering of 1 mM EGTA. b, Simultaneous [Ca²⁺] measurement (dark trace) deduced from 10 μM Fluo-4FF Ca²⁺-indicator fluorescence from entirety of same cell (514 nm excitation, 545 nm emission). Ca²⁺ determination rendered ratiometric by inclusion of 2.5 μM red Alexa 568 dye (514 nm excitation, 580 nm long-pass emission). Inset, cartoon of spatial Ca²⁺ gradients in this cell. Red traces, projected [Ca²⁺] waveforms from Ca²⁺ diffusion mechanism constrained by Ca²⁺ input in a and aggregate [Ca²⁺] (dark trace): lower red trace, aggregate [Ca²⁺]; upper red trace, submembranous [Ca²⁺]. c, Average FRET-ratio (R_F/C) response of Ca_V2.2/TN-XL channels in TIRF volume (green trace); same protocol as in a and b. Red trace, output of TN-XL forward transform in e, when driven by submembranous [Ca²⁺] input (b, upper red trace). d, Normalized steady-state FRET-ratio response (R_norm = (R_F/C − R_min)/(R_max − R_min) of Ca_V2.2/TN-XL to [Ca²⁺] (experimental green symbols). Red trace, steady-state output of TN-XL forward transform in e, when driven by differing fixed [Ca²⁺]. e, State-diagram representation of Ca_V2.2/TN-XL. Red circles, Ca²⁺ ions. f, System of states and corresponding fluorescence/FRET properties constitutes a forward transform that predicts TN-XL sensor outputs as a function of time-varying Ca²⁺ inputs.

Fig. 4e displays the corresponding state-diagram approximation of Ca_V2.2/TN-XL responsiveness to Ca²⁺, based upon a mechanism for an older troponin-based GECI⁴⁸ (TN-L15). State UB₀ (‘unbound’) represents the Ca²⁺-free conformation of the sensor, characterized by a lower FRET-ratio output of R_min. Upon binding of a single Ca²⁺ to the N-lobe of troponin, the sensor adopts one of two alternate ‘bound’ conformations (B₁ or B₂), both featuring the same elevated FRET-ratio output of R_mid. The arrangement thus far is identical to that previously established for TN-L15⁴⁸. The improved performance of TN-XL arises via engineering the C-lobe of troponin to dynamically bind and unbind two signaling Ca²⁺ ions, thereby driving a further conformational change of the sensor⁴⁴. To account for this feature, we allowed two additional Ca²⁺ ions to bind and induce a third bound state^50,51 (B₃, dashed box), exhibiting the highest FRET-ratio output of R_max. Numerical simulations of this scheme (Supplementary Methods), coupled with error minimization via parameter variation, yielded impressive fits to the target constraints above (Figs. 4c and 4d, red curves), using the sensor parameter estimates in Table 1. Accordingly, these parameters and the scheme in Fig. 4e furnish a steady-state and dynamic representation of Ca_V2.2/TN-XL responsiveness to Ca²⁺, a ‘forward transform’ (Fig. 4f) potentially appropriate for making inferences about channel nanodomain Ca²⁺ fluctuations.

Table 1.

Ca_V2.2/TN-XL parameters for HEK293 cells under TIRF imaging

Fluorescence Parameters1	Value
CaV2.2/TNXL Rmin	1.10 (n = 20 cells, Fig. 2f)
CaV2.2/TNXL Rmid	1.532
CaV2.2/TNXL Rmax	2.10 (n = 15 cells, Fig. 2f)
CaV2.2/TNXL α0	2.00 (n = 8 cells)
CaV2.2/TNXL α1	1.573

Kinetic Parameters4	Value
k01	174.4 ms−1 M−1
k10	0.00239 ms−1
k12	0.00174 ms−1
k21	0.00309 ms−1
k23	0.126 × 106 ms−1 M−2
k32	1/733 ms−1

¹Fluorescence parameters obtained following experimental procedures used to characterize TN-L15⁴⁸, unless specified otherwise.

^2,4Set by global fitting of Fig. 4e scheme to data Figs. 4a, 4c, and 4d.

³Linear interpolation on R values. Explicitly, α₁ = α₀ − (R_mid − R_min)/(R_max − R_min).

Ca_V2.2/TN-XL sensors respond to nanodomain Ca²⁺ signals

This transform could only be utilized, however, if a substantial fraction of Ca_V2.2/TN-XL sensors respond to channel nanodomain Ca²⁺ signals, as specifically produced by individual channels fluxing Ca²⁺ into their own nanodomain. Thus far, we had only demonstrated that Ca_V2.2/TN-XL sensors in the TIRF volume respond well to a generalized increase in submembranous [Ca²⁺], but this would occur whether the majority of channel-sensor fusion constructs are properly trafficked to the surface membrane, or still plentiful in submembranous vesicles within the TIRF volume. To address this issue, we again undertook simultaneous recordings of whole-cell Ca²⁺ current and TIRF imaging of Ca_V2.2/TN-XL sensors, but here with high intracellular Ca²⁺ buffering present to restrict Ca²⁺ elevations to the nanodomain of active channels that flux Ca²⁺. Specifically, we dialyzed cells with 10 mM EGTA, yielding a Ca²⁺ nanodomain radius of ~40 nm¹³ (Fig. 5), only two-fold larger than electron micrographic estimates of Ca²⁺ channel diameters³². Additionally, we only investigated cells with modest Ca²⁺ currents (~1 nA with 10 mM Ca²⁺ as charge carrier) and little Ca²⁺-dependent inactivation²¹ (CDI), so as to exclude Ca²⁺ buffer depletion. Fig. 5a summarizes the results of an exemplar cell satisfying this criterion, with the displayed data evoked by a single voltage pulse. Clearly present is a transient decrease in CFP fluorescence (S_C, cyan trace), accompanied by a corresponding phasic elevation of the FRET ratio R_F/C (green trace). Exponential fits (solid black curves) revealed dominant rise and fall times of 880 and 2000 ms for S_C, and 880 and 400 ms for R_F/C. Both the return of TN-XL waveforms towards baseline, and the modest CDI of channels under elevated Ca²⁺ buffering²¹ (Fig. 5a compared to Fig. 4a), argue against unintended depletion of Ca²⁺ buffer. Accordingly, these results suggest responsiveness of Ca_V2.2/TN-XL sensors to genuine channel nanodomain Ca²⁺ signals. By contrast to the exemplar, about half the cells with currents of this magnitude and limited CDI exhibited no appreciable change of TN-XL readouts, presumably due to poor trafficking of active channels. Such cells were excluded from further analysis.

Figure 5. Ca_V2.2/TN-XL sensors respond to channel nanodomain Ca²⁺ signals.

Top cartoon, Ca²⁺ nanodomains under elevated Ca²⁺ buffering, where Ca²⁺ elevation is restricted to nanodomains, each affiliated with individual channel. 40 nm nanodomain calculated with Equation 1, f = 1 (classical assumption), unitary current i = 0.085 pA, and conservatively requiring [Ca²⁺] ~60 nM (the approximate resting physiological Ca²⁺ concentration). a, Simultaneous records of whole-cell current (top) and TIRF fluorescence signals (middle and bottom) from HEK293 cell expressing Ca_V2.2/TN-XL, using 10 mM external Ca²⁺ and 10 mM internal EGTA Ca²⁺ buffering. Shown are data from a single trial, evoked by the first step depolarization in the cell. Top, Voltage and current waveforms, with moderately-sized current chosen to minimize buffer consumption. Middle, Corresponding CFP fluorescence signal after normalization to baseline level (S_C, noisy trace). Black curve, approximate exponential fits (initial phase: 200 ms (10%), 880 ms (90%); recovery phase: 2000 ms (100%)). Vertical ticks, 0.1 increment. Bottom, Corresponding FRET-ratio signal (R_F/C, noisy trace). Black curve, approximate exponential fits (initial phase: 200 ms (10%), 880 ms (90%); recovery phase: 400 msec (70%), 2000 ms (30%)). Vertical ticks, 0.1 increment. b, Average Ca_V2.2/TN-XL responses to nanodomain Ca²⁺ signals, as isolated with 10 mM internal EGTA Ca²⁺ buffering. Protocol and format as in panel a. n = 4 cells, with records taken from the first step depolarization delivered in cells. CDI sometimes increased with repeated depolarizations, reflecting buffer depletion with attendant slowing of R_F/C waveform decay. Cells exhibiting peak currents ≤500 pA were not considered, as these cells demonstrated negligible R_F/C transients, presumably reflecting insufficient trafficking of active channels. Solid-black-curve exponential fits were similar, with S_C exponential fit (initial phase: 200 ms (10%), 880 ms (90%); recovery phase: 500 ms (80%), 2500 ms (20%)), and R_F/C fit (initial phase: 200 ms (10%), 880 ms (90%); recovery phase: 400 msec (70%), 2000 ms (30%)). Dashed black curve, continuation of initial phase S_C exponential fit, which asymptotes at lower horizontal black line in middle. c, Schematic of various classes of Ca_V2.2/TN-XL species in TIRF volume. Blue shading indicates region of excitation by evanscent wave in TIRF volume.

Forward transform estimates of nanodomain Ca²⁺ amplitudes

As a prelude to more rigorous assessment of Ca_V2.2/TN-XL responses to nanodomain signals, we averaged signals from multiple responsive cells dialyzed with 10 mM EGTA. The mean responses (Fig. 5b) were similar to those of the exemplar, with nearly identical characteristic time constants. This reproducibility encouraged us to quantitatively scrutinize these averaged waveforms. In particular, while the robust responses observed here under elevated Ca²⁺ buffering do indicate an appreciable fraction of active channels within the TIRF volume (Fig. 5c, active), the data do not exclude the possibility of a still substantial fraction of channels that are present in the plasmalemma but fail to open (Fig. 5c, silent), or resident within submembranous vesicles within the TIRF volume (Fig. 5c, intracellular). Such silent and/or intracellular channels would contribute static background fluorescence that could complicate quantitative interpretation. An indication of this scenario comes from exponential extrapolation of the S_C waveform decline (Fig. 5b, dashed curve), which asymptotes in the range of ~0.8. By contrast, under the assumption that all channels are active, simulations of the forward transform in Figs. 4e and 4f indicate that, over a large range of plausible nanodomain Ca²⁺ pulse amplitudes (≥ 35 μM), S_C waveforms would asymptote near Ca²⁺-saturating levels of ~0.5 (Supplementary Figure S2, Supplementary Methods), as specified by the experimentally determined 1/α₀ value (defined in Table 1). Hence, the experimentally deduced asymptote of ~0.8 (Fig. 5b) suggests that ~40% of Ca_V2.2/TN-XL channels are active within the TIRF volume, while ~60% of channels are intracellular or silent. Cognizant of this configuration, we correct S_C and R_F/C waveforms to reflect only active channels, simply by subtracting ~60% of the baseline amplitude of CFP (S_C) and FRET (S_F) signals, and taking a ratio of these subtracted signals to obtain a corrected R_F/C. Fig. 6a shows the corrected signals (S̄_C, R̄_F/C), averaged over multiple cells. These waveforms were then suitable for making inferences about nanodomain Ca²⁺.

Figure 6. Estimates of nanodomain Ca²⁺ amplitude from Ca_V2.2/TN-XL responses.

a, Average Ca_V2.2/TN-XL responses to nanodomain Ca²⁺ signals, corrected to reflect signal of active channels alone. Top, black trace plots average whole-current, rescaled to open probability P_O by normalizing peak current during 30-mV pulse to 0.69. Gray trace, exponential fit to P_O waveform. Middle, cyan trace plots average S̄_C, corrected CFP signal. Red trace, forward transform prediction of experimental S̄_C using Ca_spike = 60 μM. Bottom, Green trace plots average R̄_F/C, corrected FRET-ratio signal. Red trace, forward transform prediction of experimental R̄_F/C using Ca_spike = 60 μM. Gray traces, transform prediction with Ca_spike as labeled. b, Nanodomain Ca²⁺ signals input to forward transform for prediction of Ca_V2.2/TN-XL responses in a. Top, P_O waveform reproduced from a. Middle, Schematic of stochastic Ca²⁺ pulses in nanodomain. Bottom, Forward transform box with parameters in Table 1. c, Ca_spike estimation criteria. Presumed Ca_spike values plotted versus fold-fitting error (sum of squared deviations between fit and data, normalized to sum with Ca_spike = 60 μM). Cusp of plot denotes Ca_spike = 60 μM as best fit. Shaded zone specifies approximate confidence region, where fits yield errors within 25% of minimum.

That nanodomain Ca²⁺ signals take the form of equi-amplitude Ca²⁺ pulses synchronized to the millisecond opening of channels is widely accepted^10,12 (cartoon, Fig. 6b); however, the critical amplitude of such pulses (Ca_spike) has only been inferred via simulations of Ca²⁺ diffusion^12–16. To estimate Ca_spike empirically, we analyzed the corrected Ca_V2.2/TN-XL responses (Fig. 6a) using the forward transform in Figs. 4e and 4f. The experimental waveforms might initially appear difficult to conceptualize, since each TN-XL sees a different stochastic record of Ca²⁺ pulses driven by single-channel openings. However, because the response of each TN-XL is slow (Table 1, compare Figs. 4b, 4c) relative to the millisecond duration of individual Ca²⁺ spikes, we can adopt a previously established kinetic simplification¹⁰ that permits the Ca²⁺-driven forward reaction rates in Fig. 4e to be treated, not as stochastic entities that fluctuate with each Ca²⁺ pulse, but as smoothly changing entities k₀₁·Ca_spike·P_O(t) and k₂₃·Ca²_spike·P_O(t). Importantly, these entities are to the first order equal for all active Ca_V2.2/TN-XL sensors, such that the collective output from multiple sensors, as displayed in Fig. 6a, would approximate that for any individual sensor on average. Thus, using our forward transform (Figs. 4e and 4f) to predict experimental waveforms becomes simple. Single-channel data (Fig. 3e) permit average whole-cell currents (Fig. 5b) to be rescaled as an open probability P_O waveform (Figs. 6a and 6b); this P_O drives straightforward numerical integration of the differential equations corresponding to the transform in Fig. 4e; and this integration yields predictions of experimental waveforms, with the magnitude of Ca_spike being the only unknown (Supplementary Methods). While presumed Ca_spike values of 15 or 240 μM yield predicted R̄_F/C trajectories (Fig. 6a, gray smooth curves) that are patently inconsistent with the experimental curve, Ca_spike = 60 μM produces an impressive fit to data (smooth red curves). Indeed, error analysis (Fig. 6c) confirms 60 μM as a best estimate of Ca_spike. As a conservative lower limit, we performed transform analysis on uncorrected waveforms (Fig. 5b), yielding a still-large estimated Ca_spike ~15 μM (Supplementary Figure S3, Supplementary Methods). Fits to uncorrected data were notably inferior, further prompting us to favor Ca_spike estimates from the background-corrected data in Fig. 6a. For these data, the shading in Fig. 6c encompasses fits with errors similar to the minimum, yielding an approximate Ca_spike range of 45–80 μM. This estimate, combined with determination of the underlying single-channel current amplitude, can potentially furnish valuable insight into Ca²⁺ diffusion properties within the nanodomain itself (Fig. 7).

Figure 7. Potential fenestrated Ca²⁺ egress within Ca_V2.2 channel nanodomain.

a, On-cell single-channel ramp experiments, except 10 mM Ca²⁺ is charge carrier to accord with whole-cell experiments; 700-Hz lowpass filtering. Voltages adjusted to align single-channel activation with whole-cell activation curves. Exemplars from single patch, with openings to GHK relation (gray curve). Unitary current i at 30 mV (green circles), 0.085 pA in 5 patches. b, Gain factor A, slope of Ca_spike versus i relation. Shading, various estimates of A over confidence range of Ca_spike. Sloped dashed line, lower-limit estimate of A using uncorrected data. c, Schematic, FRET construct to estimate r_sensor. Substituting circularly permuted amber for circularly permuted Citrine of TN-XL maintains sensor-channel configuration. d, Estimating r_sensor via FRET. Each black symbol, individual HEK293 cell expressing construct in c. Two outputs for each such cell obtained via 3³-FRET analysis⁵⁴ of fluorescence images: FRET efficiency and ratio of donor (CFP) to acceptor (YFP) molecules (D/A ratio), both plotted versus CFP intensity (∝ construct expression level). Data from cells exhibiting D/A ratio ~1 (as expected from c) are displayed and analyzed (bottom subpanel, black line at unity). Invariant FRET efficiency with increasing CFP intensity (top subpanel) excludes spurious FRET, facilitating interpretation of FRET efficiency (~0.35, blue line, top subpanel, left axis) as indicative of r_sensor ~55 Å (top subpanel, right axis). Gray symbols, control cells expressing YFP and CFP separately, yielding efficiencies ~0. e, A versus sensor distance from channel mouth (r_sensor), calculated with Neher-Stern equation (Eq. 1), using parameters^13,60 D_Ca = 0.4 μm² ms⁻¹ and k_Bon = 10⁻³μM⁻¹ ms⁻¹ for EGTA. Gray curves, plots with f values as labeled. Mapping A from b, with r_sensor from d, yields f ~0.53 (black curve) and range (0.4–0.7, shading). f, Schematic, idealized fenestrated Ca²⁺ egress from inner channel mouth. Green cones, permissive radial diffusion paths from source (green dot), summing to f ~ 0.53. Target molecule (gray ball) within/near portal enjoys preferential signaling.

DISCUSSION

The fusion of Ca_V2.2 channels to the genetically encoded Ca²⁺ sensor TN-XL here furnishes a new tool for detecting submembranous Ca²⁺ signals near these ‘N-type’ Ca²⁺ channels (Fig. 1b). Our particular application concerns the nanodomain Ca²⁺ transients accompanying individual channel openings and closings. These transients have eluded direct experimental comment until the recent advent of TIRF microscopy to image single-channel activity^30,31, and the availability of targetable chemical-fluorescent Ca²⁺ indicators³⁸. Here we demonstrate an alternative approach exploiting TIRF/patch-clamp electrophysiology and our Ca_V2.2/TN-XL fusion. This tactic yields a first empirical estimate of the amplitude of nanodomain Ca²⁺ pulses (Ca_spike). This entity figures crucially in the Ca²⁺ activation of co-localized channels and enzymes², evoked release of nearby neurotransmitter vesicles²⁵, and excitation-transcription coupling^8,9.

A limitation of the present study concerns the incomplete targeting of active Ca_V2.2/TN-XL channels to the surface membrane in the TIRF volume, necessitating approximate correction of static background fluorescence. The potential error in such correction renders our Ca_spike determination as a coarse approximate estimate. Nonetheless, the approach taken will prove valuable as means to improve channel trafficking arise, and GECI technology progresses.

With this proviso, we consider other long-sought-after Ca²⁺ channel signaling properties that could, in principle, be accessed with our strategy. First, we estimate the gain factor A, defined as the ratio of Ca_spike to unitary current i fluxing through individual Ca²⁺ channels¹². As such, the magnitude of A is crucial for local Ca²⁺ channel signaling to downstream nanodomain targets. Given knowledge of Ca_spike, we could estimate A itself, so long as the i were known under the conditions of Ca_spike determination (Fig. 6, with 10 mM Ca²⁺ as the charge carrier). The single-channel data thus far (Figs. 3c, 3d) were obtained with 90 mM Ba²⁺ to facilitate resolution of open probability; these data could not be utilized here. The relevant single-channel data for parameter A specification would entail several-fold smaller signals^25,52. Nonetheless, despite signals near the limits of detection, we acquired well-resolved unitary currents under these conditions (Fig. 7a). These data yield i = 0.085 ± 0.003 pA (n = 5 patches) at 30 mV (as in Fig. 6), resembling values recorded in native Ca_V2.2 channels²⁵. Combining i with our estimates of Ca_spike (center ~60 μM; range 45–80 μM), we obtain A ~700 μM/pA (Fig. 7b, shaded regions).

Second, if the distance from the cytoplasmic mouth of the channel to the sensor (r_sensor) were also known, nearfield Ca²⁺ sensing could comment powerfully on Ca²⁺ diffusion within the nanodomain itself. We therefore gauge r_sensor from FRET measured between the CFP of our Ca_V2.2/TN-XL fusion, and YFP fused to the base of the channel amino terminus (Fig. 7c). To eliminate crosstalk, YFP within TN-XL itself is replaced with the inert analog amber⁵³. FRET efficiency was thus determined⁵⁴ (Fig. 7d), yielding an estimated r_sensor ~55 Å (Methods).

With both parameter A and r_sensor in hand, we scrutinize nanodomain Ca²⁺ diffusion via classic point-source Ca²⁺ diffusion, as given by a generalized Neher-Stern equation^13,16

$$A=\frac{{Ca}_{\text{spike}}}{i}=\frac{(1/f)}{4·\pi ·F·D_{\text{Ca}}·r_{\text{sensor}}}·exp\left[\frac{-r_{\text{sensor}}}{\sqrt{D_{\text{Ca}}/(k_{\text{Bon}}·B_{\text{T}})}}\right]$$ (1)

where r_sensor is the distance of the TN-XL sensor from the point source of Ca²⁺ influx, D_Ca is the diffusion coefficient of free Ca²⁺, F is Faraday’s constant, k_Bon is the on rate for Ca²⁺ binding to EGTA, and B_T is the internal EGTA concentration (10 mM). The parameter f, specifying the fraction of half-infinite space into which radial Ca²⁺ diffusion occurs, has always been set at unity (Ca²⁺ diffuses into a full half-infinite space), absent any evidence to the contrary. Intriguingly, our estimates of A and r_sensor square poorly with the traditional view (Fig. 7e, dashed curve with f = 1), but fit well with f ~ 0.53 (0.53 × 2π steradians, solid curve). This outcome may represent an early functional indication of fenestrated Ca²⁺ egress from the channel into cytoplasm (Fig. 7f), according with crystallographic structures of K channels⁵⁵.

Thus, nanodomain Ca²⁺ diffusion could focus Ca²⁺ through particular molecular geometries within the channel-signaling complex, raising the possibility of enhanced preferential signaling to target molecules near or within exit portals (Fig. 7f, gray ball). Indeed, while prior studies elegantly suggest that distinctive unitary current amplitudes i render Ca_V2.2 channels favorable for triggering vesicle fusion²⁵, it is also plausible that differing A and f values also factor into such optimization. The latter possibilities represent an intriguing, but nearly unexplored realm of Ca²⁺ biology. Though other interpretations are certainly viable, further empirical estimates of Ca_spike, parameter A, and r_sensor will likely aid in this field of inquiry.

Methods

Molecular biology

Ca_V2.2/TN-XL (Ca_V2.2 is of human origin (NM000718)) fusion construct was made by PCR, using fusion primers. Pfu polymerase (Strategene) was used for fidelity. To make Ca_V2.2/TN-XL, we first generated a Ca_V2.2 channel with the C-terminus truncated at the 2180th amino acid and without the stop codon, Ca_V2.2Δ2180 stopless/XbaI in pcDNA3.1. This was made by cutting Ca_V2.2 with Xho I and Xba I, and replacing the resulting fragment with the PCR product amplified from Ca_V2.2 between the Xho I and 2180th amino acid sites using forward primer CGCATCAGTTACAATGACATG, and reverse primer CTGTCTAGAAGCACCAGATGTTGA CAGCA. Next, TN-XL was PCR-amplified with an Xba I site upstream of CFP (ATG removed) and a Nhe I/Spe I site downstream of the stop codon of Citrine. The forward and reverse primers used were CGGCCGCCACCTCTAGAGTGAGC, and CAGACTAGTGCTAGCTTAGTCCTC GATGTTGTGGC respectively. Finally, the TN-XL PCR fragment was spliced into the Ca_V2.2Δ2180 stopless/XbaI construct at the Xba I site to give our Ca_V2.2/TN-XL construct. For the variant construct in Fig. 7c, the circularly permuted Citrine in TN-XL was replaced with circularly permuted amber⁵³, and YFP fused to the amino terminus of Ca_V2.2 after removal of the first 81 amino acids²² (nearly the entire amino terminus).

Verification by complete sequence analysis of regions subjected to PCR was performed.

HEK293 cell culture and transfection

HEK293 cells were cultured on No-1 25 mm glass cover slips (Bellco glass) in 10 cm plates. These slips were coated overnight with 0.01% (wt/v; diluted 1:5 in 0.1 M Borate buffer) poly-D-lysine (Sigma), and washed with ddH₂0 on the day cells were seeded onto the slips. The poly-D-lysine coating facilitates proper whole-cell voltage control at the membrane/glass interface⁵⁶, and also increases electrostatic attraction between the glass surface and cells, thus favoring cell attachment and optimal TIRF imaging. Cells were transiently transfected, using calcium-phosphate precipitation⁵⁴, with Ca_V2.2/TN-XL, rat brain β_2a⁵⁴, and rat brain α_2δ⁵⁴ (10 μg each). Ca_V2.2 and β_2a-CFP were substituted for Ca_V2.2/TN-XL and β_2a respectively in some experiments. Cotransfection of channels with a number of molecules did not appreciably improve channel trafficking in our system: PI3K and Akt/PKB⁵⁷, ORL receptors⁵⁸, and dominant-negative dynamin⁵⁹. TN-XL and TN-XL-Ras sensors were transfected using FuGENE6 (Roche).

Electrophysiology

Whole-cell and single-channel current records were obtained at room temperature 1–3 d post-transfection using patch-clamp amplifiers Axopatch 200B and Axopatch 200A (Axon Instruments), respectively.

For whole-cell recordings, the cells were bathed in external solution containing (in mM): TEA-MeSO₃, 140; HEPES (pH 7.4 with TEA-OH), 10; and CaCl₂, 10; at 300 mOsm, adjusted with glucose. The internal solution contained (in mM): Cs-MeSO₃, 114–135; CsCl₂, 5; EGTA, 1 or 10; MgCl₂, 1; MgATP, 4; and HEPES (pH 7.4 with CsOH), 10; 295 mOsm with Cs-MeSO₃. In some experiments, 10 μM Fluo 4FF and 2.5 μM Alexa 568 (Invitrogen, Molecular Probes) were included for ratiometric Ca²⁺ determination. Electrodes were pulled from borosilicate glass capillaries(WPI MTW 150-F4) and had pipette resistances ranging from 1.5–2.5 MΩ before 75% series resistance compensation. Voltage pulses were applied at 90 sec intervals. Currents were filtered at 5 kHz, and digitized at 25 kHz. Leak and capacitance transients subtracted by P/8 protocol. Data acquired and analyzed with custom MATLAB scripts (Mathworks, Natick, MA).

Single-channel recordings were all conducted in the on-cell configuration. The bath contained (in mM): K Glutamate, 132; KCl, 5; NaCl, 5; MgCl₂, 3; EGTA, 2; glucose, 10; and HEPES (pH 7.4 adjusted with KOH), 20; at 300 mOsm adjusted with glucose. This bath solution zeroed the membrane potential. The pipette solution contained, (in mM): BaCl₂, 90; TEA-MeSO₃, 20; HEPES (pH 7.4 adjusted with TEA-OH), 10; at 300 mOsm, adjusted with TEA-MeSO₃. In some experiments, 10 mM CaCl₂ was substituted for BaCl₂ and TEA-MeSO₃ correspondingly increased. To minimize electrical noise, 5–20 MΩ patch pipettes were pulled from ultra-thick-walled (2 mm outer diameter, 1.16 mm inner diameter) borosilicate glass (BF200-116-10, Sutter Instruments), and coated with Sylgard. Seal resistance was ~80 GΩ. Voltage pulses applied at repetition intervals of 5–15 s; data sampled at 70 μsec intervals, and filtered at 2 kHz (Fig. 3c, d) or 700 Hz (Fig. 7a).

TIRF microscopy

On the day of experiments, the cover slips on which cells were cultured were mounted on a custom bath chamber for TIRF imaging. Fluorescence measurements of single cells were performed on a Nikon Eclipse TE2000U microscope, equipped with a Plan Apo 60 × TIRF objective (NA 1.45 oil immersion). Nikon immersion oil, with nd = 1.515 (23 °C) was used. For TN-XL imaging, CFP excitation was delivered by a 16-mW Melles Griot 56ICS 442-nm diode laser. For Fluo 4FF/Alexa 568 imaging, excitation was delivered by a 40-mW Melles Griot IMA100 argon laser featuring a 514-nm line. For TN-XL imaging, the optical path diagrammed in Supplementary Figure S4 was used, as described below. A high-resolution CoolSNAP HQ CCD camera (Photometrics), driven by IPLab (Scanalytics) scientific imaging software, was used for cell orientation and verification under TIRF microscopy. Fast quantitative data, averaged across single cells, were instead acquired via PMTs, as follows. Single-cell fluorescent signals was selected by the image-plane pinhole in a D-104C dual-channel photometer (PTI), and sent to a dichroic mirror (510dclp) which divides the fluorescence emission signals into CFP and FRET signals. These two signals passed through their respective emitters (HQ485/30m and HQ545/40m, Chroma), before being captured by two Hamamatsu R1527 PMTs. Fluorescent signals were filtered at 2 kHz and sampled at 25 kHz, and reflect the activity of several thousand Ca_V2.2/TN-XL molecules. Shutter control, dark current subtraction, and data acquisition were controlled by custom MATLAB programs. For Fluo 4FF/Alexa 568 imaging, the optical path was modified slightly. Single-cell fluorescence was isolated via the same pinhole and then sent to a dichroic mirror (545 dclp) which parsed signals into Fluo 4FF and Alexa 568 signals. These two signals passed through their respective emitters (545/40 nm bandpass and HQ580 nm lp, Chroma), before being captured by the same PMTs. The apparent K_d for Fluo 4FF/Alexa 568 ratios under our recording conditions was 20 μM, as determined in live HEK293 cells dialyzed with Ca²⁺ dye and a series of calibrated [Ca²⁺] internal solutions.

Widefield FRET microscopy

3³-FRET efficiency determination, along with estimation of the ratio of donor to acceptor molecules, was performed within individual live HEK293 cells imaged through a image-plane pinhole according to CFP, FRET, and YFP filter cubes⁵⁴. Moreover, r_sensor in Fig. 7d was estimated using a Forster distance of 49 Å⁵⁴.

Immunoblots

HEK293 cells expressing Ca_V2.2/TN-XL or TN-XL-Ras were washed and harvested with PBS, centrifuged, and resuspended in lysis buffer (20 mM Tris-HCl [pH 7.5], 300 mM sucrose, 1 mM DTT and protease inhibitor cocktail (Roche)). Lysis was completed with freeze-thaw cycles and sonication. Membrane proteins were separated by initial centrifugation (16,000g ×10 min at 4°C), followed by solubilizing the resulting pellet in 1% Triton X-100. This solution was then subjected to centrifugation (16,000g ×10 min at 4°C), yielding membrane proteins in the supernatant. Membrane proteins were denatured (95°C ×1 min) upon dilution with TGS buffer (Bio-Rad), and resolved by 5% SDS-PAGE with no CaCl₂ added to the gel. Proteins were transferred to nitrocellulose membranes (Bio-Rad) by cooled transfer (Tris/Glycine transfer buffer/pH 8.3 ×19 h). After blocking, membranes were sequentially incubated with rabbit polyclonal anti-GFP antibody (Abcam Inc., 1:5000 dilution as directed by manufacturer) and HRP-conjugated secondary antibody (Sigma, 1:2400 dilution as recommended by manufacturer). Protein bands were visualized with enhanced chemiluminescence (Pierce ECL, Thermo Scientific).

Data Analysis and Modeling

Single-channel ramp data were analyzed for Ca_V2.2/TN-XL unitary current and open probability using custom MATLAB programs¹⁰. For the analysis, all patches contained 1 or 2 active channels. TIRF fluorescent signals from Ca_V2.2/TN-XL were modeled using custom MATLAB scripts, using approaches detailed in Supplementary Methods. Generalization of the Neher-Stern equation (Eq. 1) for f values other than unity was accomplished by considering mass balance and substituting Q₁/f for Q₁ in Eq. 6 of the original Stern paper¹³.